IJE Advance Access originally published online on September 19, 2006
International Journal of Epidemiology 2006 35(5):1151-1159; doi:10.1093/ije/dyl185
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Point-Counterpoint |
The population dynamics of cancer: a Darwinian perspective
1 Chair of Environmental Epidemiology, Department of Epidemiology and Public Health, Imperial College London, St Mary's Campus, Norfolk Place, W2 1PG, London, UK
2 University of New Mexico, Department of Internal Medicine, New Mexico Cancer Research Facility, MSC08 4630, Room 103A, 1 University of New Mexico, Albuquerque, NM 87131, USA
* Corresponding author. E-mail: p.vineis{at}imperial.ac.uk
 |
Abstract |
|---|
 Top Abstract Epidemiological evidenceSpecific examples of...Mutations and selectionNon-linear dynamics and the...Conclusions and perspectivesReferences |
|---|
Carcinogenesis, at least for some types of cancer, can be interpreted as the consequence of selection of mutated cells similar to what, in the theory of evolution, occurs at the population level. Instead of considering a population of organisms, we can refer to a population of cells belonging to multicellular organisms. Many carcinogens are mutagens, and the observed geographic distribution of cancer is, at least in part, attributable to environmental mutagens. However, the rapid change in risk for some cancers after migration suggests that carcinogenesis involves—in addition to mutations—some late event that most probably consists of the selection of cells already carrying mutations. We review a few examples of such selective pressures: finasteride in prostate cancer, vitamin supplementation in smokers, acquired resistance to chemotherapy, peripheral resistance to insulin, and sunlight and mutations in melanoma. A disease model for such a hypothesis is represented by Paroxysmal Nocturnal Hemoglobinuria (PNH). Mutations can be present at birth, as in the case of PNH, and can have a frequency much higher than the occurrence of the corresponding disease (PNH or lymphocytic leukaemia in children). However, PNH does not require a mutator phenotype, only a mutant phenotype followed by selection. A characteristic feature of cancer, instead, is likely to be the development of the mutator phenotype. We propose a ‘Darwinian’ model of carcinogenesis. If the model is correct, it suggests that prevention is more complex than avoiding exposure to mutagens. Mutations and genetic instability can be already present at birth. Mutations can be selected in the course of life if they increase survival advantage of the cell under certain environmental circumstances. In addition, gene–environment interactions cannot be interpreted according to a simplified linear model (based on the ‘analysis of variance’ concept); experimental work suggests that a more comprehensive non-linear interpretation based on the idea of ‘norm of reaction’ is needed.
Accepted 23 July 2006
The purpose of this paper is to revisit the role of mutations in carcinogenesis, suggesting that cancer can arise not only as effect of a ‘mutant phenotype’, i.e. the comparison of mutated cells with aberrant behaviour, but actually of a ‘mutator phenotype’, i.e. of cells that are particularly prone to multiple mutations. In addition, we propose that selective pressure over mutants and, even more, mutator cells, is likely to be crucial in explaining the role of environmental exposures. We start with the description of some epidemiological examples: the descriptive epidemiology of some types of cancer, suggesting that rapid changes in cancer incidence can occur after migration or lifestyle changes; and examples of potential ‘selectogenic’ environmental exposures. By environmental exposures we mean all changes that occur in the external environment that have the capability of modifying also the internal environment of the cell. Therefore, when we refer to the effect of migration, we hypothesize that it affects cancer rates because of material changes (e.g. in dietary habits) that in turn modify the internal environment of cells and thus contribute to the selection of mutated cells particularly adapted to those environments.
|
Epidemiological evidence |
|---|
|
TopAbstractEpidemiological evidenceSpecific examples of...Mutations and selectionNon-linear dynamics and the...Conclusions and perspectivesReferences |
|---|
Cancer incidence is extremely variable across the world.1 Such variation is not attributable to genetic characteristics of different populations, for at least three reasons: (i) highly penetrant mutations explain only 5–10% of all cancers, while geographic variation is—for instance, for melanoma or oesophageal cancer—up to 1:200; (ii) low-penetrant gene variants (polymorphisms) are only weakly associated with cancer risk; and (iii) for most cancers, the risk of migrants tends to approach that of the population to which they migrate, sometimes as soon as in the first generation. Therefore, variation in cancer incidence must be largely associated with environmental exposures. For a long time mutations have been considered to be the main mechanism by which cancer arises; or, more accurately, they have been considered to be necessary but not sufficient for carcinogenesis,2 but the role of selection of mutated cell has not been clearly investigated. Mutations are not likely to be the only mechanism to explain geographic variation and incidence changes in migrants. An alternative model is mutation plus selective pressure over mutated cell clones.
One can look at the geography of cancer as Darwin studied the phenotypes of different species in his journey around the world. As the selection of certain genotypes was favourable in specific environments and not in others, similarly, the selection of cells carrying certain mutations can lead to a greater incidence of cancer under specific environmental circumstances. We will show a few examples that corroborate this hypothesis.
The geography of cancers of the digestive tract
Cancers of the digestive tract show a peculiar geographic distribution that sheds light on possible aetiology. Colorectal malignancies, on one hand, and oesophageal, gastric, and liver malignancies, on the other hand, show in fact opposite patterns of geographic distribution.
For colorectal cancer there is a 25-fold variation in occurrence worldwide, with very high incidence rates in North America, Australia, and Europe (annual incidence rates 
60 per 105 or higher), while incidence is quite low in Africa and Asia (Table 1). (Japan is in fact an exception, because it has now the highest incidence in the world, and the change has been very recent). The opposite happens for stomach cancer, with two out of three cases per 105 per year occurring in developing countries, very high rates in Asia, and moderately high rates in Africa.
|
Oesophageal and liver cancers are almost entirely diseases of developing countries, where more than 80% of the cases occur, although oesophageal cancer is now rising rapidly in Western countries. The highest rates of oesophageal cancer are found in Asia and, in particular, in China (more than 180 new cases per 105 per year), while for liver cancer the highest occurrence is in Africa. Liver cancer is clearly a viral disease, being largely explained by hepatitis B and C. Incidence patterns on a world scale seem to follow the patterns of infection: where hepatitis B is high, liver cancer is high, and vaccination in Taiwan led to reduced incidence. Increasing incidence of hepatitis C also seems to be associated with increasing incidence of liver cancer in the US and Japan. Risk factors for oesophageal cancer seem to be more disparate.
Finally, pancreatic cancer, the aetiology of which is totally obscure, does not show a clear international pattern.
Migrants: effects in the first and in the following generations
The effect of migration is dramatic for certain types of cancer, namely colorectal cancer and melanoma, with the incidence increasing rapidly after migration from low-risk to high-risk areas. Both increase dramatically within the first generation of migrants.3–6
In British migrants to Australia,7 a clear gradient of increasing risk was observed with duration of stay in Australia, particularly for colorectal cancer. The same pattern for colorectal cancer was observed in the Japanese who migrated to Hawaii. Conversely, migrant populations from high-risk areas of the world show a marked diminution in risk for stomach and liver cancer when they move to a lower risk area. This phenomenon seems to depend on the age at migration, and childhood environment is important in determining risk.
In the case of gastric and liver cancers, exposure to Helicobacter pylori (stomach cancer) or to hepatitis B or C viruses (liver cancer) in childhood would set the risk typical of the areas of origin, while reduced exposure to cofactors might help to explain the lower risk after migration. In the case of liver cancer, a reduction in exposure to aflatoxins is believed to explain reduction in risk after migration. Cofactors can be interpreted as ‘selectogens’, i.e. exposures that facilitate selection of mutated cells (see below).
The effect of migration is much more evident in the second and the following generations of migrants. Rates of stomach, liver, and oesophageal cancer decrease rapidly in subsequent generations of migrants from high-risk to low-risk areas, while rates of colorectal cancer increase from low-risk to high-risk areas of the world.8
The migrant studies of course could not distinguish between mutagenic vs non-mutagenic effects. The sense of our reasoning is that changes among migrants were so rapid that it is unlikely that they can be due to early events in carcinogenesis, such as mutations that are fixed in all subsequent cell generations, but are more likely to be due to modulating events that act later in the carcinogenic process (a concept with a long tradition, starting with Armitage and Doll's model of carcinogenesis, see below).
Time trends in cancer mortality in China
It is also relevant to observe what happens in a rapidly developing society like China. Although incidence data are not available, extensive, good quality mortality data can be used. These reflect both aetiology and quality of care (survival), but show changes that are perfectly compatible with inferences drawn from the experience of international variation and patterns in migrants: a rapid decrease for both oesophageal and stomach cancers, and an increase for colorectal cancer. Data for liver cancer are more ambiguous. More generally, cancers related to infectious agents are declining (e.g. cervix uteri) while cancers typical of wealthy societies are on the increase (e.g. breast).9
|
Specific examples of ‘selectogens’ |
|---|
|
TopAbstractEpidemiological evidenceSpecific examples of...Mutations and selectionNon-linear dynamics and the...Conclusions and perspectivesReferences |
|---|
Vitamin supplementation in smokers and lung cancer
Paradoxical results were obtained in randomized controlled trials when heavy smokers were treated with high doses of vitamins (mainly retinol and carotenoids). Cancer occurrence was in fact higher in vitamin-treated subjects, at least in the first years after the conclusion of trials (which in some cases was stopped earlier for ethical reasons).10 No clear explanation has been put forward for this observation. However, a rather obvious explanation that comes to mind is the possibility that certain cells carry mutations, induced by tobacco smoke, that confer a selective advantage in the presence of vitamins. Vitamins have generally beneficial and protective effects towards several chronic diseases, which have been attributed for example to their anti-oxidant properties,11 but they also induce mitosis through an action on cell-cycle progression.12 The presence of mutations combined with vitamin administration could further enhance mitosis in the mutated cells in comparison with normal cells. This mechanism can help explain the paradoxical results of trials in smokers, as well as the inconsistent results of studies on supplementation with micronutrients, e.g. flavonoids.
Hyperinsulinaemia, peripheral resistance to insulin and cancer
There is increasing evidence that cancers at different sites (in particular of the breast, colon, pancreas, and prostate) are associated with mechanisms that include hyperinsulinaemia, peripheral resistance to insulin and increased production of IGF-1.13 These mechanisms are not directly mutagenic. A possible explanation for the increased risk of cancer is that such mechanisms increase the proportion of cells that undergo mitosis (mitogenic effect). In fact, insulin has several actions including regulation of cell growth, differentiation, and metabolism. These varied biological effects of insulin result from the activation of a wide array of intracellular signalling proteins involved in multiple post-receptor pathways. Binding of insulin to its receptor (IR) activates a tyrosine kinase; such signalling, in turn, branches off into two pathways that lead to either the activation of ERK1/2 or of phosphatidylinositol 3-kinase (PI 3-kinase). Activation of either pathway has been implicated for the mitogenic effect of insulin in different cell types, whereas metabolic responses elicited by insulin are more closely linked to the PI 3-kinase pathway. Mutated cells in the colon or pancreas (and, in particular, cells with specific mutations) may be more sensitive than others to the pro-mitogenic activity of insulin and this would explain the observed association between hyperinsulinaemia and cancer.
Prostate cancer and anti-androgenic therapy
The international variation in prostate cancer is largely a function of the use of PSA as a screening test and of early detection practices. However, it is interesting to observe that international variation is largely due to invasive, infiltrating forms of prostate cancer. The prevalence of latent cancer (detectable with PSA) shows much less geographic variation than clinical cancer, where information on PSA patterns is available. This observation is interesting in the light of recent evidence on the ability of anti-androgen therapy (finasteride) to cause an increase of more aggressive, infiltrating cancers. In a randomized trial, although finasteride overall prevented prostate cancer (with a 25% reduction in the experimental group), the frequency of Gleason grades 7 or more was higher in the finasteride-treated group (37%) than in the control group (22%, P < 0.001).14
The most obvious explanation is that finasteride selects more malignant cells, which would have a selective advantage in conditions of androgen inhibition. Geographic variation of prostate cancer—which is almost entirely due to infiltrating, high grade tumours—might be due to environmental agents/behaviours that interfere with androgen metabolism and select more aggressive cells. Unfortunately the epidemiology of prostate cancer is very little understood. Although the available evidence suggests an involvement of androgens as a risk factor, a distinction between rapidly progressive and clinically more benign tumours has not been clearly made yet in epidemiological studies.15
Sunlight and skin cancer
Different types of skin cancers have different relationships with sunlight. Squamous-cell skin cancer is considered to be strongly associated with chronic, prolonged exposure to sunlight. The association between melanoma and sunlight is weaker and more complex, and in fact short-term and intense exposure seems to be more effective. Basal cell carcinoma seems to be intermediate in terms of the pattern and dose of sunlight that is associated with risk.16 Recently, evidence has been published on the frequency of mutations in skin cancers. In particular, melanoma cells shows mutations in BRAF genes but only in sun-exposed sites in contrast to melanomas arising in non-sun exposed sites.17,18 The authors of the studies stress that BRAF mutations are likely to be relevant to early phases of tumour initiation. In fact, BRAF mutations are present in the majority of naevi (benign precursors of melanomas). However, these mutations were unlikely to be due to sun exposure, since the C to T transitions that are a characteristic fingerprint of sun exposure in melanomas (UVB signature) were not observed. It was, therefore, suggested that the concentration of BRAF mutations in sun-exposed areas was due to an indirect mechanism of sunlight, possibly by selecting already mutated cells.17,18 It should be noted that BRAF mutations can generate an immune response, but this is late in melanoma progression, usually many years after the mutation has appeared.19–21 This is further evidence that the mutation is an early event, which in itself does not lead to melanoma until further (sun-related) events occur. Such events would lead to both melanoma progression and the appearance of immunogenicity of BRAF, possibly through a mechanism of selection of cells bearing BRAF mutations.
Resistance to chemotherapy
A recent study has shown that lung cancer cells acquire resistance to the chemotherapeutic agents Gefitinib or Erlotinib by developing a double mutation in the epidermal growth factor receptor (EGFR) gene.22 A first, gain-of function mutation was already known to confer increased sensitivity to the drugs. The second mutation in the same gene, on the other hand, confers resistance to the drugs and, in fact, a selective advantage to the cells that carry it. Treatment with Gefitinib or Erlotinib allows double mutated (resistant) clones, that otherwise would be extremely rare, to become apparent. While the probability for the double mutation in the same gene to occur spontaneously or by chance alone is extremely low, selection of the mutated clone makes the double mutation a frequent occurrence.
In fact, resistance to chemotherapy23 and to carcinogens24 is now interpreted more in terms of gross changes in chromosome number (aneuploidy), or chromosome aberrations (instability), rather than of point mutations. Genetic variability allowed by varying chromosome numbers or increased instability is in fact much greater and allows more flexible adaptation to sudden environmental stresses.
|
Mutations and selection |
|---|
|
TopAbstractEpidemiological evidenceSpecific examples of...Mutations and selectionNon-linear dynamics and the...Conclusions and perspectivesReferences |
|---|
Somatic mutations (not present in the germline cells) can be already present at birth. A striking recent observation was the finding of a very high proportion, in healthy newborns, of fusion genes TEL-AML1 and AML-ETO associated with lymphocytic leukaemia (the mutation rate was about 100 times higher than the cumulative incidence of leukaemia).25 While the origin of such mutations is not known—but could be associated with in utero exposure to genotoxicants—it is clear that mutations per se are insufficient to explain the onset of leukaemia, which is probably due to further ‘hits’ that select cells with a selective advantage (an additional mechanism, epigenetic events, will be addressed below). In another investigation in humans, Finette et al. 26 found a high prevalence of hprt mutations at birth in healthy children, coming to similar conclusions as Mori and colleagues. In a series of well-designed experiments, Somers et al.27 reported increased mutation rates in herring gulls and mice exposed to air pollution at levels that characterize normal urban environments. In mice, in fact, mutations were transmitted transgenerationally, i.e. they were attributed to DNA damage in sperm cells. Somatic mutations in newborns have been related to air pollutants,28 and mutations in germ cells have been attributed to air pollution or cigarette smoking.29,30
Fetuses and newborns seem to be particularly susceptible to carcinogens. In one study, mother–newborn pairs exposed to high levels of outdoor air pollution were investigated.31 For all markers, including DNA adducts, newborns had levels that were higher than those in the mothers, although transplacental exposure levels were 10 times lower than the paired mother exposures. In a second study from the same group, environmental prenatal exposure to Polycyclic Aromatic Hydrocarbons resulted in higher frequencies of chromosome aberrations in the cord blood of newborns.32 In one experiment, pregnant rats were exposed to environmental tobacco smoke; 8-OH-dG adducts were formed in the fetal kidney, liver, and brain, with increases that were similar to the risk of human cancer related to ETS. The distribution in different organs depended on gestational stage.33 Perinatal dietary habits and dietary restriction influence the formation of adducts from endogenous oxidative compounds. In rats subjected to dietary restriction, hepatic DNA adducts due to oxidative DNA stress were significantly decreased compared with a parallel ad libitum fed control group.34 Therefore, it is not unlikely that DNA damage is already present at birth, both due to direct mutagenic stimuli and to indirect mechanisms, thus predisposing to cancer if further hits occur.
Certain mutations can confer, in particular circumstances, a selective advantage, or can work as ‘adaptive mutations’. Unfortunately there are not many examples of human diseases that correspond to this model. However, at least one example of how the presence of adaptive mutations can confer advantage in stressful circumstances is Paroxysmal Nocturnal Hemoglobinuria (PNH), which is an acquired stem cell disorder characterized by intravascular haemolysis, hypercoagulability, and bone marrow failure. The characteristic defect in paroxysmal nocturnal haemoglobinuria is the somatic mutation of the PIG-A gene in haematopoietic cells. The current hypothesis explaining the disorder suggests that there are two components: (i) haematopoietic stem cells with the characteristic defect are present in the marrow of many normal individuals in very small numbers; (ii) some aplastogenic influence (e.g. an immune overreaction to drugs) suppresses the normal stem cells but does not suppress the defective stem cells, thus, allowing the proportion of these cells to increase (a ‘Darwinian’ interpretation).35 Those who are not carriers of PIG-A mutations and are exposed to the aplastogenic insult instead of developing PNH develop an even worse disease, aplastic anaemia. Of course PNH and aplastic anaemia can have strongly diverging incidence rates in the population, depending on the frequency of exposure to aplastogenic agents. Why the majority of normal, healthy individual carry PIG-A mutations in blood cells is totally unclear.
PNH can serve as a model of a disease owing to a common mutation present at birth that remains silent unless a second hit (in this case an immune or autoimmune reaction) intervenes.
Cell proliferation or cell selection?
It is frequently (and incorrectly) supposed that cancer is characterized by increased cell proliferation. In fact, according to pathologists, there are three possible mechanisms by which cells can accumulate: (i) cancer cells replicate faster; (ii) a smaller proportion die, or (iii) a greater proportion of daughter cells replicate compared with normal cells. In fact, the two first mechanisms are not observed frequently by pathologists. The cell cycle time, even for the fastest-growing human tumours like acute leukaemia, is often longer than for normal cells.36 Increased mitosis rate is not considered to be a constant feature of cancer, and cell death is a frequent observation. Thus, pathologists suggest that the most common explanation for cell accumulation in cancer is a greater proportion of daughter cells replicating. This is considered to be the effect of selective pressure exerted over daughter cells after they have undergone a mutation.37
Fisher's F-fitness and mathematical models of carcinogenesis
Carcinogenesis, at least for some types of cancer, can thus be interpreted as the consequence of selection of mutated cells similar to what, in the theory of evolution, occurs at the population level. Instead of considering a population of organisms, we can refer to a population of cells belonging to multicellular organisms.
Ronald Fisher described in a mathematical form the relationship between fitness and time variables (‘Fisherian fitness’)38:
 |
It has been shown38 that F-fitness is appropriate in a population with exponential growth (typically a culture of bacteria or daughter cells multiplying from a stem cell). This is the case in which the effect of Darwinian selection is observed more evidently (‘survival of the fittest’). When growth is less than exponential, Michod et al.38 have shown that ‘any’ phenotype can evolve, i.e. new types can increase regardless of their individual capacities (‘fitness of anybody’). Finally, for a population growth that is greater than exponential, the expression ‘fitness of the first’ has been coined, meaning that individual capacities do not count, and the common types persist over time because benefits must be proportional to the square of density (for mathematical formulations see Michod, ref. 38).
Gatenby39 has proposed a more complex model that is derived from Fisher's equations and describes the relationships of two populations, in this case normal cells and malignant cells, competing for space and other resources:
 | 1 |
 | 2 |
ij = coefficient of competition that measures the effects of population i on population j. In the initial phase of growth of a cancer, approximately N1 << K1 (so that N1/K1 << 1) and N2/K2 approximately = 1.
Therefore, Equations (1) and (2) become:
 | 3 |
 | 4 |
In the invasive phase, when N2 << N1, Equations (1) and (2) become
 | 5 |
 | 6 |
12 = 0). These equations suggest that (i) the initial phases of malignant transformation involve competition between normal cells and cancer cells, i.e. carcinogenesis can be counteracted by normal cells if these have sufficient resources (i.e. prevention could consist in empowering normal cells); (ii) the final, invasive phase involves changes that enable the cancer cells to definitely overcome normal cells, mainly by angiogenesis and the adoption of the glycolytic phenotype.40
This model strongly resembles the mathematical model developed by Moolgavkar et al.41 called two-stage clonal expansion (TSCE) model of carcinogenesis, which is being increasingly used for the analysis of epidemiological data. However, a crucial difference is that the TSCE model does not assume competing cell populations, i.e. cell populations are statistically independent. The model simulates the initiation, promotion, and transformation processes of normal and initiated cell populations during the subject's life. The outcome of the model is the probability of appearance of the first malignant cell. Normal cells can mutate into intermediate cells with a Poisson process of intensity 
X, where is the first mutation rate. Then, mutated cells can replicate with rate , die with rate ß, or mutate into malignant cells with a second mutation rate µ. The first mutational stage corresponds to the emergence of the malignant phenotype, the ratio between replication and death rates represents proliferation of transformed cells, and the second mutation leads to predominance of the invasive phenotype. Therefore, Moolgavkar's model is compatible with Gatenby's equations, which, however, consider the normal cells' ability to compete with cancer cells. A generalization of Moolgavkar's model has been proposed by Little and Wright.42
|
Non-linear dynamics and the origins of the mutator phenotype |
|---|
|
TopAbstractEpidemiological evidenceSpecific examples of...Mutations and selectionNon-linear dynamics and the...Conclusions and perspectivesReferences |
|---|
Breivik in a recent paper43 has raised two important points: (i) most models of carcinogenesis have been based on a linear combination of events (e.g. mutations), but the process of carcinogenesis is likely to be highly non-linear, as the accumulation of mutations and chromosome damage in cancer cells suggests; (ii) random mutations are unlikely to lead to a selective advantage of cells; for simple statistical reasons, the vast majority will be neutral or will lead to cell death. Point (i) is in fact confirmed by the previous paragraph on mathematical models, because all of them start with a linear approach. Point (ii) implies that there is some additional mechanism other than random mutations induced by carcinogenic exposures, i.e. some mutations would have to cause a ‘cascade’ of events, according to the idea of the ‘mutator phenotype’. Breivik attributes this cascade of events to something that affects the efficiency of DNA repair when the rate of mutations is too high. In fact, in a situation of stress the selective advantage would go to cells that do not to stop to repair damage, i.e. ‘fine-tuning’ would lead cells into a waste of energy and time and these would not survive.
The same issue has been raised when people studied the ability of bacteria to survive in the absence of certain nutrients. Also in the latter case, the ability that bacteria acquired to metabolize different substrates could not be explained by random mutations.
Hypermutability in bacteria as response to stress
In the first observations of the ability of bacteria to overcome environmental stress, it was noted that Escherichia coli in the stationary phase yielded adaptive mutations at rates beyond what might be expected from their frequency per cell division in growing cultures. A strain with a deletion of a gene for ß-galactosidase, when incubated for days in the presence of lactose, developed another enzyme for the utilization of that substrate.44 In fact, it became clear that the appearance of such mutants required two mutations, one in the gene (ebg) for the new enzyme and one in its repressor. The product of the two mutation rates was far too low to account for the double mutations. The presence of lactose seemed to stimulate their appearance, but this sounded like an unacceptable ‘lamarckian’ perspective. A closer look showed that there was a burst of mutations in the ebg gene, providing a large population within which the second mutation then provided an occasional adaptive clone. The similarity with the two mutations required to acquire resistance to chemotherapeutic agents is noteworthy.
More generally, bacteria under environmental stress undergo a condition called SLAM or stressful lifestyle associated mutation (or stationary-phase mutability). This condition requires functional proteins of the double-strand break repair recombination system, i.e. recombination is part of a mechanism by which stationary-phase mutations form. In conditions of stress mutations occur at a rate that is 10–1000 times higher than that in control bacteria populations.45,46 However, mutations do not occur in all cells, but only in surviving subpopulations, which leave the hypermutable state when the adaptive mutation(s) is generated47 (i.e. once they are able to survive they are no longer hyper-mutable). Survivors will all carry the mutation. In this transition Mismatch Repair (MMR) plays a central role, being inhibited during the stationary-phase.48,49 Similar modulation in stationary-phase has been shown in human cells.49 Among the conditions that trigger the stationary-phase there are starvation, and physical and chemical stress. The small minority of heritable mutator mutants resembles that seen in several examples of adaptive evolution. Interestingly, the selected stress-induced mutator alleles are positively correlated with the strength of selection and negatively with the frequency of such stresses.
At least for one environmental contaminant, cadmium, at the dose levels typical of human exposure, mutagenesis has been shown to be indirect and related to inhibition of MMR. Jin et al.50 found that chronic exposure of yeast to environmentally relevant concentrations of cadmium resulted in extreme hypermutability. In extracts of human cells cadmium inhibited mismatch removal.
However, it is likely that in carcinogenesis, rather than point mutations, more macroscopic changes are involved, including, in particular, changes in chromosome numbers and chromosome instability.23,24 Genetic variability allowed by varying chromosome numbers or increased instability is in fact much greater and allows more flexible adaptation to sudden environmental stresses.
Norms of reaction and canalization
Finally, it is worth mentioning an important misunderstanding about the role of genes in causing disease. This issue was raised in an important paper by Richard Lewontin51 many years ago, but it is still matter of confusion. The main idea of the paper was that when we talk about genes, environment, and gene–environment interactions we use the ‘analysis of variance’ paradigm, i.e. we try to combine the two main effects (genes vs environment), plus their interactive term, in a linear model. Variances are then computed and the role of the two main effects (or their interaction) is apportioned accordingly. But this is totally misleading. There is no reason to use a linear combination: this is done for the sake of simplicity but does not correspond to any reasonable biological state assumption. In contrast, all the experiments done with e.g. Arabidopsis (a plant) or Drosophila (based for example on radiation-induced mutations) show that mutations cause a change in what is called the ‘reaction norm’, i.e. the ability of the organism to react to changing environmental conditions. The reaction norm was originally defined as ‘the function that relates the environments to which a particular genotype is exposed and the phenotypes that can be produced by that genotype’.52 The phenotype is not predictable according to a linear function that relates the genotype with the environmental conditions. For example, in average environmental conditions (e.g. temperature) there is no difference between the ‘wildtype’ and the mutant genotypes, i.e. the phenotype is the same. However, hidden genetic variation is revealed when environmental conditions change and deviate from average.
In fact, we can imagine at least two different situations: (i) one in which the environmental stressor leads to a certain dose–response curve in those with a gene variant, and a steeper dose–response relationship for another gene variant (e.g. cholesterol levels in relation to dietary habits in subjects with two different variants of a hypercholesterolaemia gene); (ii) in other situations, the mutation/variant may be necessary but not sufficient for observing the effect, i.e. in those with the wildtype no effect is observed (these issues have been discussed at length by Ottman, ref. 53). The case of ‘reaction norm’ is closer to type (i) above, because it involves different quantitative phenotypic responses as a consequence of variant genotypes.54 The typical example is the one put forward by Lewontin himself, i.e. the ability of plants to adapt to different altitudes, a quantitative trait, depending on their genotypes: it is not a question of all or none response but rather of modulation of the response in different environments. A related concept, also coined by Waddingon, is that of ‘canalization’: since organisms have hidden genetic variability, which is revealed only under stressful conditions, repeated exposure of different generations will lead to the selection of the variant genotypes. Under normal circumstances variant genotypes are allowed to accumulate because they are not ‘seen’ by natural selection (i.e. the phenotypic response is ‘canalized’ towards the usual response), but when environmental stressors appear repeatedly, the reaction norm is canalized towards a different phenotype through the selection of the variant genotypes. A very clear introduction to many of the evolutionary concepts exposed in this paper can be found in the book by Jablonka and Lamb.55 Most of these concepts have not been incorporated in cancer research as yet.
Epigenetics: an additional layer of complexity
Our hypothesis suggests new avenues for epidemiological research but leaves some open questions. For example, in addition to mutation and selection, what is the role for DNA methylation patterns or other epigenetic events? DNA methylation, that is the covalent addition of methyl groups (–CH3) to cytosine that precedes a guanosine in the DNA sequence (the CpG dinucleotide), occurs naturally and is thought to have a role in suppressing gene expression. This is called an epigenetic modification, because it does not change the structure of DNA, but is heritable, and passes from one generation of cells to the next. Hypermethylation of promoter regions is associated with gene transcriptional silencing and is a common mechanism for the inactivation of tumour suppressor genes in human cancer. In elegant experiments in mice, nutritional changes during pregnancy were shown to interfere with the subsequent cancer risk through methylation patterns. The agouti vs yellow colour of the mice hair is determined by methylation patterns. If the agouti gene terminal repeat region is hypermethylated, the mouse is agouti, if it is hypomethylated the mouse is yellow. When pregnant mice were fed a diet rich in folate and methionine (i.e. in methyl groups), none of the pups born were yellow. Interestingly, the expression of the yellow coat, as a consequence of folate deficiency and hypomethylation, was linked to an increased risk of obesity, adult diabetes, cancer, and mortality.56,57 In other words, intrauterine exposure to nutrients associated with epigenetic modifications of the genome in the offspring can lead to increased cancer risk. This adds a further level of complexity to our hypothesis.
|
Conclusions and perspectives |
|---|
|
TopAbstractEpidemiological evidenceSpecific examples of...Mutations and selectionNon-linear dynamics and the...Conclusions and perspectivesReferences |
|---|
Many carcinogens are mutagens, and the observed geographic distribution of cancer is, at least in part, attributable to environmental mutagens. However, the rapid change in risk for cancers of the digestive tract after migration, the effect of anti-androgens in prostate cancer, and other epidemiological examples suggest that carcinogenesis involves—in addition to mutations—some late event that most probably consists of the selection of cells already carrying mutations. A disease model for such hypothesis is represented by PNH. However, PNH does not require a mutator phenotype, only a mutant phenotype followed by selection. A characteristic feature of cancer is likely to be the development of the mutator phenotype. Some somatic mutations can be present at birth, as in the case of PNH, and can have a frequency much higher than the occurrence of the corresponding disease itself (PNH or lymphocytic leukaemia in children).
The model we propose distinguishes itself from the standard initiation–promotion model simply because it puts emphasis on its coherence with our background biological knowledge, which is essentially based on the evolutionary theory. ‘Promotion’ is not some mysterious property of some chemicals, but is circumstantial on the ability of cells carrying specific mutations to have selective advantage in a specific environment. Our idea of course is not new and has been put forward from disparate points of view, for example by Charlton (‘endogenous parasitism’),58 Greaves (from a more clinical point of view),59 or Weiss, from a more strictly genetic point of view.60
If this model is correct, it suggests that prevention is more complex than avoiding exposure to mutagens. Although this is certainly a necessary step, we have to consider that some mutations can be already present at birth and can be selected in the course of life if they increase survival advantage of the cell under certain environmental circumstances. In addition, gene–environment interactions cannot be interpreted according to a simplified linear model (based on the ‘analysis of variance’ concept); experimental work suggests that a more comprehensive non-linear interpretation based on the idea of ‘norm of reaction’ is needed. Of course, complex exposures can play a role as both mutagens and selectogens. This is likely to be the case for tobacco smoking and for air pollution (see for example ref. 61), which contain mutagens but could also act by preferentially selecting already mutated cells.
|
Acknowledgments |
|---|
This work has been made possible by grants from the World Cancer Research Fund and Compagnia di San Paolo (Torino) to P.V. We are grateful to Jarle Breivik, John Potter, and Mark Little for critical comments.
|
References |
|---|
|
TopAbstractEpidemiological evidenceSpecific examples of...Mutations and selectionNon-linear dynamics and the...Conclusions and perspectivesReferences |
|---|
1 Parkin DM, Bray F, Ferlay J, Pisani P. Global cancer statistics, 2002. CA Cancer J Clin 2005;55:74–108.
2 Cahill DP, Kinzler KW, Vogelstein B, Lengauer C. Genetic instability and darwinian selection in tumours. Trends Cell Biol 1999;9:M57–60.[CrossRef][Web of Science][Medline]
3 Iscovich J, Howe GR. Cancer incidence patterns (1972–91) among migrants from the Soviet Union to Israel. Cancer Causes Control 1998;9:29–36.[CrossRef][Web of Science][Medline]
4 Iscovich J, Parkin DM. Risk of cancer in migrants and their descendants in Israel: I. Leukaemias and lymphomas. Int J Cancer 1997;70:649–53.[CrossRef][Web of Science][Medline]
5 Swerdlow AJ, Marmot MG, Grulich AE, Head J. Cancer mortality in Indian and British ethnic immigrants from the Indian subcontinent to England and Wales. Br J Cancer 1995;72:1312–19.[Web of Science][Medline]
6 Mack TM, Floderus B. Malignant melanoma risk by nativity, place of residence at diagnosis and age at migration. Cancer Causes Control 1991;2:401–11.[CrossRef][Web of Science][Medline]
7 McCredie M, Williams S, Coates M. Cancer mortality in migrants from the British Isles and continental Europe to New South Wales, Australia, 1975–1995. Int J Cancer 1999;83:179–85.[CrossRef][Web of Science][Medline]
8 Thomas DB, Karagas MR. Cancer in first and second generation Americans. Cancer Res 1987;47:5771–76.
9 Yang L, Parkin DM, Ferlay J, Li L, Chen Y. Estimates of cancer incidence in China for 2000 and projections for 2005. Cancer Epidemiol Biomarkers Prev 2005;14:243–50.
10 Goodman GE, Thornquist MD, Balmes J et al. The beta-carotene and retinol efficacy trial: incidence of lung cancer and cardiovascular disease mortality during 6-year follow-up after stopping beta-carotene and retinol supplements. J Natl Cancer Inst 2004;96:1743–50.
11 IARC Working Group on the Evaluation of Cancer Preventive Agents. IARC handbooks of cancer prevention. Vol. 2, Carotenoids. Lyon, France: IARC Scientific Publications, 1998.
12 Bohnsack BL, Hirschi KK. Nutrient regulation of cell cycle progression. Annu Rev Nutr 2004;24:433–53.[CrossRef][Web of Science][Medline]
13 Barnard RJ, Aronson WJ. Preclinical models relevant to diet, exercise, and cancer risk. Recent Results Cancer Res 2005;166:47–61.[Medline]
14 Reynolds T. Prostate cancer prevention trial yields positive results, but with a few cautions. J Natl Cancer Inst 2003;95:1030–01.
15 Makridakis NM, Buchanan G, Tilley W, Reichardt JK. Androgen metabolic genes in prostate cancer predisposition and progression. Front Biosci 2005;10:2892–903.[Web of Science][Medline]
16 Kricker A, Armstrong BK, English DR, Heenan PJ. Does intermittent sun exposure cause basal cell carcinoma? a case-control study in Western Australia. Int J Cancer 1995;60:489–94.[Web of Science][Medline]
17 Maldonado JL, Fridlyand J, Patel H et al. Determinants of BRAF mutations in primary melanomas. J Natl Cancer Inst 2003;95:1878–90.
18 Cohen Y, Rosenbaum E, Begum S et al. Exon 15 BRAF mutations are uncommon in melanomas arising in nonsun-exposed sites. Clin Cancer Res 2004;10:3444–47.
19 Andersen MH, Fensterle J, Ugurel S et al. Immunogenicity of constitutively active V599EBRaf. Cancer Res 2004;64:5456–60.
20 Fensterle J, Becker JC, Potapenko T et al. B-Raf specific antibody responses in melanoma patients. BMC Cancer 2004;4:62.[CrossRef][Medline]
21 Sharkey MS, Lizee G, Gonzales MI, Patel S, Topalian SL. CD4(+) T-cell recognition of mutated B-RAF in melanoma patients harboring the V599E mutation. Cancer Res 2004;64:1595–99.
22 Pao W, Miller VA, Politi KA et al. Acquired resistance of lung adenocarcinomas to gefitinib or erlotinib is associated with a second mutation in the EGFR kinase domain. PLoS Med 2005;2:e73.[CrossRef][Medline]
23 Li R, Hehlman R, Sachs R, Duesberg P. Chromosomal alterations cause the high rates and wide ranges of drug resistance in cancer cells. Cancer Genet Cytogenet 2005;163:44–56.[CrossRef][Web of Science][Medline]
24 Bardelli A, Cahill DP, Lederer G et al. Carcinogen-specific induction of genetic instability. Proc Natl Acad Sci USA 2001;98:5770–75.
25 Mori H, Colman SM, Xiao Z et al. Chromosome translocations and covert leukemic clones are generated during normal fetal development. Proc Natl Acad Sci USA 2002;99:8242–47.
26 Finette BA, Homans AC, Rivers J, Messier T, Albertini RJ. Accumulation of somatic mutations in proliferating T cell clones from children treated for leukemia. Leukemia 2001;15:1898–905.[Web of Science][Medline]
27 Somers CM, Yauk CL, White PA, Parfett CL, Quinn JS. Air pollution induces heritable DNA mutations. Proc Natl Acad Sci USA 2002;99:15904–07.
28 Perera F, Hemminki K, Jedrychowski W et al. In utero DNA damage from environmental pollution is associated with somatic gene mutation in newborns. Cancer Epidemiol Biomarkers Prev 2002;11:1134–37.
29 Selevan SG, Borkovec L, Slott VL et al. Semen quality and reproductive health of young Czech men exposed to seasonal air pollution. Environ Health Perspect 2000;108:887–94.[Web of Science][Medline]
30 DeMarini DM. Genotoxicity of tobacco smoke and tobacco smoke condensate: a review. Mutat Res 2004;567:447–74.[CrossRef][Web of Science][Medline]
31 Whyatt RM, Jedrychowski W, Hemminki K et al. Biomarkers of polycyclic aromatic hydrocarbon-DNA damage and cigarette smoke exposures in paired maternal and newborn blood samples as a measure of differential susceptibility. Cancer Epidemiol Biomarkers Prev 2001;10:581–88.
32 Bocskay KA, Tang D, Orjuela MA, Liu X, Warburton DP, Perera FP. Chromosomal aberrations in cord blood are associated with prenatal exposure to carcinogenic polycyclic aromatic hydrocarbons. Cancer Epidemiol Biomarkers Prev 2005;14:506–11.
33 Maciag A, Bialkowska A, Espiritu I et al. Gestation stage-specific oxidative deoxyribonucleic acid damage from sidestream smoke in pregnant rats and their fetuses. Arch Environ Health 2003;58:238–44.[Web of Science][Medline]
34 Zhou G, Hernandez NS, Randerath E, Randerath K. Effects of different diets and dietary restriction on perinatal endogenous DNA adducts. Time dependence of oxidative and presumptive nonoxidative lesions. Mutat Res 2000;447:137–47.[Web of Science][Medline]
35 Bessler M, Schaefer A, Keller P. Paroxysmal nocturnal hemoglobinuria: insights from recent advances in molecular biology. Transfus Med Rev 2001;15:255–67.[Web of Science][Medline]
36 Tannock I. Cell kinetics and chemotherapy: a critical review. Cancer Treat Rep 1978;62:1117–33.[Web of Science][Medline]
37 Fischer AH, Young KA, DeLellis RA. Incorporating pathologists' criteria of malignancy into the evolutionary model for cancer development. J Cell Biochem 2004;93:28–36.[CrossRef][Web of Science][Medline]
38 Michod R. Darwinian Dynamics. Princeton: Princeton University Press, 2000.
39 Gatenby RA. Application of competition theory to tumour growth: implications for tumour biology and treatment. Eur J Cancer 1996;32A:722–26.
40 Gatenby RA, Vincent TL. An evolutionary model of carcinogenesis. Cancer Res 2003;63:6212–20.
41 Moolgavkar SH, Luebeck EG. Multistage carcinogenesis and the incidence of human cancer. Genes Chromosomes Cancer 2003;38:302–06.[CrossRef][Web of Science][Medline]
42 Little MP, Wright EG. A stochastic carcinogenesis model incorporating genomic instability fitted to colon cancer data. Math Biosci 2003;183:111–34.[CrossRef][Web of Science][Medline]
43 Breivik J. The evolutionary origin of genetic instability in cancer development. Semin Cancer Biol 2005;15:51–60.[CrossRef][Web of Science][Medline]
44 Davis BD. Transcriptional bias: a non-Lamarckian mechanism for substrate-induced mutations. Proc Natl Acad Sci USA 1989;86:5005–09.
45 Rosenberg SM, Thulin C, Harris RS: Transient and heritable mutators in adaptive evolution in the lab and in nature. Genetics 1998;148:1559–66.
46 Bjedov I, Tenaillon O, Gerard B et al. Stress-induced mutagenesis in bacteria. Science 2003;300:1404–09.
47 Hall BG. Spontaneous point mutations that occur more often when advantageous than when neutral. Genetics 1990;126:5–16.[Abstract]
48 Harris RS, Feng G, Ross KJ et al. Mismatch repair is diminished during stationary-phase mutation. Mutat Res 1999;437:51–60.[CrossRef][Web of Science][Medline]
49 Harris RS, Feng G, Ross KJ et al. Mismatch repair protein MutL becomes limiting during stationary-phase mutation. Genes Dev 1997;11:2426–37.
50 Jin YH, Clark AB, Slebos RJC et al. Cadmium is a mutagen that acts by inhibiting mismatch repair. Nature Genetics 2003;14:326–29.
51 Lewontin R. The analysis of variance and the analysis of causes. Am J Human Genetics 1972;26:400–11. Reprinted in Int J Epidemiol June 2006.
52 Pigliucci M. Phenotypic Plasticity. Baltimore: The Johns Hopkins University Press, 2001, p. 5.
53 Ottman R. Gene-environment interaction: definitions and study designs. Prev Med 1996;25:764–70.[CrossRef][Web of Science][Medline]
54 Suzuki Y, Nijhout HF. Evolution of a polyphenism by genetic accommodation. Science 2005;311:650–52.
55 Jablonka E, Lamb MJ. Evolution in Four Dimensions. Cambridge: MIT Press, 2005.
56 Cooney C, Dave A, Wolff G. Maternal methyl supplements in mice affect epigenetic variation and DNA methylation of offspring. J Nutr 2002;138:2393S–2400S.
57 Waterland R, Jirtle R. Early nutrition, epigenetic changes at transposons and imprinted genes, and enhanced susceptibility to adult chronic diseases. Nutrition 2004;20:63–68.[CrossRef][Web of Science][Medline]
58 Charlton BG. Is immortality a possibility? A thought experiment concerning the inevitability of senescence due to endogenous parasitism. Perspect Biol Med 1995;39:146–49.[Web of Science][Medline]
59 Greaves M. Cancer causation: the Darwinian downside of past success? Lancet Oncol 2002;3:244–51.[CrossRef][Web of Science][Medline]
60 Weiss KM. Is there a paradigm shift in genetics? Lessons from the study of human diseases. Mol Phylogenet Evol 1996;5:259–65.[CrossRef][Web of Science][Medline]
61 Vineis P, Husgafvel-Pursiainen K. Air pollution and cancer: biomarker studies in human populations. Carcinogenesis 2005;26:1846–55.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  J. J.-C. Sheu, C.-H. Lee, J.-Y. Ko, G. S.W. Tsao, C.-C. Wu, C.-Y. Fang, F.-J. Tsai, C.-H. Hua, C.-L. Chen, and J.-Y. Chen Chromosome 3p12.3-p14.2 and 3q26.2-q26.32 Are Genomic Markers for Prognosis of Advanced Nasopharyngeal Carcinoma Cancer Epidemiol. Biomarkers Prev., October 1, 2009; 18(10): 2709 - 2716. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. J.-C. Sheu, C.-H. Hua, L. Wan, Y.-J. Lin, M.-T. Lai, H.-C. Tseng, N. Jinawath, M.-H. Tsai, N.-W. Chang, C.-F. Lin, et al. Functional Genomic Analysis Identified Epidermal Growth Factor Receptor Activation as the Most Common Genetic Event in Oral Squamous Cell Carcinoma Cancer Res., March 15, 2009; 69(6): 2568 - 2576. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. Odoux, H. Fohrer, T. Hoppo, L. Guzik, D. B. Stolz, D. W. Lewis, S. M. Gollin, T. C. Gamblin, D. A. Geller, and E. Lagasse A Stochastic Model for Cancer Stem Cell Origin in Metastatic Colon Cancer Cancer Res., September 1, 2008; 68(17): 6932 - 6941. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
F. Saletta, G. Matullo, M. Manuguerra, S. Arena, A. Bardelli, and P. Vineis Exposure to the Tobacco Smoke Constituent 4-Aminobiphenyl Induces Chromosomal Instability in Human Cancer Cells Cancer Res., August 1, 2007; 67(15): 7088 - 7094. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. D. Nagy, E. M. Victor, and J. H. Cropper Why don't all whales have cancer? A novel hypothesis resolving Peto's paradox Integr. Comp. Biol., August 1, 2007; 47(2): 317 - 328. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. A. Klein Random mutations, selected mutations: A PIN opens the door to new genetic landscapes PNAS, November 28, 2006; 103(48): 18033 - 18034. [Full Text] [PDF]
|
|
|
|
 |
|
 S. EBRAHIM Entelechy, citation indexes, and the association of ideas Int. J. Epidemiol., October 1, 2006; 35(5): 1117 - 1118. [Full Text] [PDF]
|
|
|
|
|
|
J. Lynch It's not easy being interdisciplinary Int. J. Epidemiol., October 1, 2006; 35(5): 1119 - 1122. [Full Text] [PDF]
|
|
|
|
|
|
E. Jablonka Commentary: Induction and selection of variations during cancer development Int. J. Epidemiol., October 1, 2006; 35(5): 1163 - 1165. [Full Text] [PDF]
|
|
|
|
|
|
K. M Weiss Commentary: Evolution of action in cells and organisms Int. J. Epidemiol., October 1, 2006; 35(5): 1159 - 1160. [Full Text] [PDF]
|
|
|
|
|
|
R. A Gatenby Commentary: Carcinogenesis as Darwinian evolution? Do the math! Int. J. Epidemiol., October 1, 2006; 35(5): 1165 - 1167. [Full Text] [PDF]
|
|
|
|
|
|
J. Breivik Commentary: Cancer--evolution within Int. J. Epidemiol., October 1, 2006; 35(5): 1161 - 1162. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||