Can Peto’s paradox be used as the null hypothesis to identify the role of evolution in natural resistance to cancer? A critical review

Underlying hypotheses of Peto’s paradox

Do cell division patterns support Peto’s paradox?

The first hypothesis of Peto’s paradox postulates that large/long-lived animals have
more dividing cells compared to smaller/short-lived ones. This hypothesis does not
take into account the great variety of division rates within an organism where some
cells could divide more frequently than others.

In many, if not most, cases, cancer may arise from transformation of stem cells 27], cells representing the first step of differentiation processes and with a great
potential to divide (and/or proliferate). During the development of multicellular
organisms, the obvious function of cell differentiation is to create new cell types.
In adult organisms, new cell types are no longer needed or produced – but cell replacement
is essential, tissues could be maintained by the self-duplication of fully mature
and functional cells. Therefore, the function of ongoing, but tightly controlled cell
differentiation may have evolved to protect from detrimental cell-level progression
28], 29]. With such a serial differentiation pattern, self-renewing cell populations are much
more susceptible to somatic mutation, but these cells are rare and slow growing. Certain
type of differentiated cells cannot initiate propagation of malignant phenotypes because
they cannot divide, e.g. myocytes, adipocytes, and neurons 30]. Based on that concept, Peto’s hypothesis assumes that the number of stem cells should
correlate with body mass. But the number of stem cells as well as the number of divisions
have a low probability to correlate with body mass. A different number of differentiated
cells may be obtained from the same number of stem cells 31] by dint of a switch between proliferation (dividing cell) and differentiation (non-dividing
cell) (Fig. 2). Then, the number of divisions will not only depend on the number of stem cells,
but also on the timing to switch between proliferation and differentiation (Fig. 2). The number of cells that will divide as well as the tissue turnover can be very
different among organs 32], for example, in humans, the intestinal epithelium completely self-renews within?~?5 days,
while lung epithelium takes up to 6 months to be replaced 28]. Furthermore the number of stem cells is also different among organs, and this number
could be involved in tumorigenesis 27]. Naively, one might think that having a larger organ requires a greater number of
cells, but recent perspective papers show that differences of cell size could also
be essential in determining organ size 33]. Including cell size as a parameter for the prediction of cancer risk shows that
the correlation between body/organ size and cancer is weaker 33]. Furthermore, basal metabolic rate (BMR) is also decreased in larger animals compared
to smaller ones (i.e. Max Kleiber allometric law 34]). Low BMR induces less oxidative stress in comparison to higher BMR 35]. Thus, larger animals could have a lower level of oxidative stress compared to smaller
ones, and hence offsetting the higher cancer risk due to increased cell numbers. Indeed,
a recent study by Dang, 2015 36] support the hypothesis that metabolism can drive tumorigenesis and accounts for Peto’s
paradox explanation.

Fig. 2. Variation in time to switch between division and differentiation results in significant
cell number differences inspite of the same starting stem cell numbers. a Stop of proliferation and start of differentiation after three generations lead to
a larger differentiated tissue mass. b Stop of proliferation after one generation and start of differentiation earlier than
A, result in a smaller differentiated tissue mass

Would transformation rates to cancer phenotypes be equivalent across different cell
and tissue types?

Another assumption of Peto’s paradox, based on the fact that the rate of malignant
transformation may be constant and similar across cell types, is that the mutation
accumulation rate is constant among the cells. Important sources of genomic alterations
are mutations, or spontaneous errors of DNA replication, 37], 38] that occur despite the existence of a wide range of mechanisms ensuring DNA repair
and correct replication 39].

However, division processes – and mutation rates – may differ among cell types: for
instance mutation rate has been reported to be 17 times higher in human somatic cells
than in germ cells 40]. The mutation rate may differ also between organs, even though there is only limited
data available on the mutational spectra of various tissue types 41]. Among differentiated cells, mutation rates of human retina cells has been estimated
to be 3.7 times greater than intestinal epithelial cells, but still 1.48 times lower
compared to that observed in lymphocytes 40] in which recombination events occur naturally and frequently. The rate of genetic
alterations also varies across species, for example for a given organ, such as colon,
mutation rates per generation is 2.14 times greater in the rat (Rattus norvegicus) than in the mouse (Mus musculus). The level of genetic variation can be intrinsic to the tissue type, e.g. the level
of oxidative stress is very different across different tissue types 35]. Furthermore, mutation rates may also be affected by exposure to mutagens, especially
in tissues, such as skin, respiratory and digestive epithelia, that are in direct
contact with the external environment and then naturally more exposed to mutagens
and radiations. Differences may also exist between similar organs in diverse animal
species 40].

As suggested above, additional mechanisms, especially for lymphomas and leukemia,
can increase DNA instability in specific cell types such as T and B lymphocytes, some
of the key cells of the vertebrate immune system. One of the important characteristics
of lymphocytes is that a specific part of their coding genome is hypermutated to generate
the incredible genetic diversity necessary to recognize the plethora of foreign antigens,
and hence protect the organism from a broad range of pathogens 42]. The enzymes involved in initiating the hypermutation events could potentially also
increase the genomic instability of these cells and favor errors leading to lymphoid
transformation 43].

Would carcinogenesis rely on mutations only?

The last assumption of Peto’s paradox is that a variety of somatic genomic alterations,
from single nucleotide variants to larger structural aberrations (including insertions,
deletions, and chromosomal translocations) can contribute to cell transformation (somatic
mutation theory 44]). The genetic alterations will then be transmitted through DNA replication and cell
division to the daughter cells. However, spontaneous mutations are insufficient to
explain cell transformation in every situation 45]–47], and cancer can potentially also arise from a variety of other mechanisms, which
may vary between organs and species (Table 1).

Table 1. Main cancer causes apart from that mentioned in Peto’s paradox

Variation in mutation numbers required to trigger tumor formation and progress

The number of genetic alterations varies largely, depending on age and tumor type,
e.g. the number of genetic alterations is usually reduced in pediatric tumors such
as juvenile myelomonocytic leukemia 48], 49] or acute megakaryoblastic leukemia 50] while being the highest in lung cancers induced by smoking 51] and melanomas induced by UV 52], 53]. The genomic signature of tumor cells (established based on the nature, localization
and number of genomic alterations identified in the affected cells) informs about
the factors that have promoted and contributed to the malignant transformation (ageing
versus toxic exposure versus genetic predisposition etc…) 54]–57]. Solid tumors usually carry more genomic alterations than hematological malignancies
58]–60].

Furthermore, the functional consequences of a given mutation are highly variable,
depending on its nature and localization in the genome. Those that have the most striking
effects are those that activate a proto-oncogene (e.g. genes involved in cancer initiation/progression)
or inactivate a tumor suppressor gene (e.g. genes that allow apoptosis or stop cell-cycle).
A single nucleotide change can be sufficient to transform a proto-oncogene into an
oncogene that induces cell transformation, whereas an inhibiting mutation must affect
the two alleles of a tumor suppressor gene to favor transformation 61].

Epigenetic factors

In addition, a growing number of studies show that epigenetic stochasticity can act
as driving force of carcinogenesis, via regulating the inhibition of tumor suppressor
genes 62] as well as the activation of proto-oncogenes 63]. Since epigenetic stochasticity is not correlated to body size, it may introduce
background noise when testing Peto’s paradox. Furthermore, since environmental factors
(e.g. species ecology, habitat, resource availability) can significantly influence
transgenerational epigenetic modifications, it can thus be important to consider both
consistent and stochastic (e.g. oil spills, famine, extreme climate parameters) environmental
changes across generations in order to decipher their contribution to tumor formation
62].

Tumor microenvironment

In addition to spontaneous mutation and epigenetic mechanisms, it is also increasingly
recognized that tissue organization plays a major role in the development of malignant
phenotypes (tissue organization field theory) 44]. This theory relies on the fact that cancer cells can proliferate only within a suitable
microenvironment 1], 64], a particular tissue environment with specific conditions, e.g. low pH and/or oxygen
concentrations 65]. Generally, normal tissue homeostasis and architecture inhibit progression of cancer,
but changes in the microenvironment can shift the balance of these signals to a cancer
permissive state. Tumor development, progression and metastasis are strongly dependent
on the microenvironmental conditions met by cancer cells 1]. Tumor ecosystems consist of non-malignant normal cells (fibroblasts, immune cells
and cells that comprise the blood vessels) and heterogeneous cancer cells, as well
as their cellular products supporting cancer cell growth. Interactions between cancer
cells and the surrounding microenvironment are constant, and bidirectional. Tumors
can influence the microenvironment by releasing extracellular signals, promoting tumor
angiogenesis and inducing peripheral immune tolerance. In return, the immune cells
in the microenvironment influence the growth and evolution of cancerous cells (e.g.
immune-editing 66]).

Animal models have demonstrated that alterations in the tissue microenvironment can
promote the emergence of clonal malignancies, e.g. mutation in Dicer genes (involved
in RNA interference) generated in the bone marrow microenvironment can promote the
emergence of a leukemic clone 67], 68]. Lastly, the recent success of immunotherapeutic strategies demonstrates that suppression
of the anticancer immune response is required for a tumor to emerge 69]. Therefore, even if cells have enough mutations to initiate carcinogenesis, malignant
cells won’t develop without a permissive cancer niche and immune system, which will
be then dependent of the tissue, the organ, and the species 65].

Thus, a Darwinian evolution of host factors relating to resistance may be more relevant
for an explanation of Peto’s paradox, than carcinogenesis parameters such as cell
divisions or stem cell number.

Peto’s paradox at the population level: artifact or reality?

Sampling bias

Assessment of Peto’s paradox 3], 4], 12], 20] relies on cancer incidence measured over very few species, i.e., dog (Canis lupus domesticus), mouse (Mus musculus), beluga (Delphinapterus leucas) and humans 12], covering a small gradient of the possible body mass. Another possible bias, when
assessing this paradox, is that the detection of cancer relates only to the presence
of macroscopic tumors, and thus neglects the precancerous lesions or the microscopic
tumors of vital organs. Thus, due to the bias of studied species, current datasets
are definitely lacking power to determine the exact relationship between body mass
and cancer incidence 70].

Additionally, other sampling biases may also explain the lack of relationship between
body mass and cancer prevalence. Of particular concern is that research so far has
predominantly relied on domesticated and laboratory animals when attempting to establish
the correlation. While the role of artificial selection for certain traits has been
recognized 71], it seems to also apply to the emergence of cancer phenotypes. Anthropogenic selection
(including domestication and breeding for particular traits in the laboratory) could
have additionally led to artificial selection for cancer resistance or susceptibility.
Therefore, laboratory and domesticated species, e.g., mice and dogs, could have cancer
incidences different from wildlife species because of an inadvertent selection of
traits involved directly or indirectly with carcinogenesis.

Environmental factors triggering the development of cancer phenotypes

Inter-species comparison can be challenging and misleading due to the fact that cancer
initiating factors are probably not the same between different species. Indeed, comparison
between human and other species could be biased by different levels and types of exposure
to environmental and behavioral factors, including pollution, abundant and excess
food supply, and frequent contact with mutagens 72], 73]. For instance, while there is no significant difference between body size of roe
deer (Capreolus capreolus) and humans (on a logarithmic scale), cancer incidence is much higher in humans (20 %
versus 2 % for roe deer) 74]–76]. These different incidences could be explained by physiological parameters, but also
by a differential exposure to mutagens. Furthermore, human cancers have been studied
more extensively and on a broader scale than the ones observed in wildlife, i.e.,
roe deer. Similarly, although extensive data is available on relatively high cancer
prevalence in Belugas (27 %), these numbers originate from a pod of whales living
in a polluted environment, suggesting that cancer prevalence could also be overestimated
for this species, just like for humans 20]. For humans, the way of life may be critically important, for instance low concordance
rate for leukemia in identical twins (5 %) suggests that additional postnatal exposure
should influence leukemia development 77].

Comparing animal species occupying different trophic levels can also jeopardize the
identification of animal species with resistance to cancer. For instance, mutation
is also driven by cellular proliferation after injuries. Therefore, species with high
injury rate from predators or aggressors should have evolved faster wound healing/tissue
regeneration 78], 79], which could concomitantly increase the number of malignant transformations due to
increased level of cell proliferation being associated with growth factors induced
in tissue regeneration 80]. Furthermore, occupying different ecological niches can also contribute to various
levels of cancer prevalence. For example, natural habitats of large mammals, such
as elephants or beluga whales (except the aforementioned pod of whales), are significantly
less polluted than the habitat of benthic organisms that are more exposed to contaminated
sediments 81].

It is recognized that for many species longevity is highly correlated with size 82], but there are also noticeable exceptions, for instance the naked-mole rat that displays
a maximum lifespan of 28.3 years for a mass of 35 g (in contrast to a similar size
Mus musculus with a maximum lifespan of 3.5 years) 83]. Due to a long-lived organism potentially accumulating more mutations during its
life 45], 84], it is expected that selection will favor cancer resistance in small, but long-lived
species to circumvent the higher risk of cancer due to mutation accumulation (e.g.
naked mole rate 85], 86]). Thus, for species displaying an atypical relationship between size and longevity,
cancer resistance pattern will not follow the traditional prediction derived from
Peto’s paradox.

Finally, increasing number of studies suggest that at least some cancers may have
infectious origins 87]. The number of pathogen known to be associated with cancer in wildlife has also been
on the rise, for example woodchucks (Marmota monax) suffering from hepatocarcinomas originating from hepatitis virus infections 88] and marine turtles succumbing to fibropapillomatosis also caused by viruses 89]. Several studies have focused on comparative analysis of parasite communities, and
on the determining factors of parasite species richness, heterogeneity and densities
90]–92]. A relationship between body size and parasite species richness is thus possible,
for example it has been shown that endogenous retroviruses abundance negatively correlates
with body mass 93].