Of barriers and loops – how evolution limits most cancer risk to older ages

Abstract

The risks and health consequences of cancers increase dramatically at older ages. To develop interventions to limit the impact of cancers, from preventative to therapeutic, we must seek both evolutionary and proximate explanations for this age-dependence. Here, we discuss how natural selection has erected barriers to delay malignancy and to maximize reproductive fitness. Each barrier need not be perfect, as long as malignant progression is delayed till older ages. With aging, mechanisms ranging from epigenetic deregulation to inflammation to senescence to mutation-driven clonal expansions contribute to increased cancer pathogenesis through mutually-enhancing mechanisms, creating tissue contexts more favorable for malignant evolution.

Why is cancer largely a disease of aging?

Most cancer biologists study cancers, the dreaded diseases that threaten our health and lives, and that of our loved ones. Still, we need to recognize that these cancers represent the extremely rare malignant successes out of the many billions of cells that set out on the potential path to cancers. Success for the individual is contingent on the failure of the vast majority of malignant clones. Indeed, as we age, our bodies become riddled with these “failures”, including billions of clonally-expanded cells with cancer-associated mutations, various abnormal growths from bronchial dysplasias to colonic polyps to skin moles, and carcinomas-in-situ in tissues such as skin, breast, prostate and thyroid(1–3). These observations raise questions: How do humans do such a good job of not developing lethal cancer for most of our lives, with about 60% of us not developing a life-threatening cancer at all? Why do some develop cancer and others not? While lifestyles and exposures (such as obesity, cigarette smoking, pollution, alcohol, and low activity) substantially contribute to the risks of cancers(4), most of these cancers are still relegated to older ages, and thus we will focus here on the key role that age plays in determining when we develop cancers, and not on extrinsic factors that increase risk.

This review will address these questions by discussing the barriers that natural selection has erected to limit the fitness-reducing impacts of malignant somatic evolution. Fitness is the relative ability of a genotype to contribute to future generations, and this simple definition applies to populations (from viruses and bacteria to blue whales) to somatic cells in our bodies. Cancers are not simple diseases, and nor are the myriad mechanisms that we and other animals have evolved to prevent them. As we will discuss below, these mechanisms need not be perfect, as the objective of natural selection is not to prevent all somatic evolution or even all malignant evolution, but just the somatic evolution that will reduce individual fitness (see Box 1). From this perspective, we can also consider how natural selection has erected tumor suppressive barriers that not only limit the more damaging somatic evolution, but which delay the negative consequences of somatic evolution until old age. A better understanding of the nature of these barriers could uncover as of yet untapped potential for improved therapeutic strategies that leverage potent cancer protective mechanisms that evolution has provided us.

Box 1. Evolutionary perspective on aging and disease risks.

To properly consider how cancer risk and pathogenesis are altered by aging, we need to first step back and delve into the underlying evolutionary theory. We will start with an evolutionary first principle that should apply to all animals – natural selection drove the development of a soma that maximizes reproductive success, and thus the maintenance of the soma wanes at ages where individuals are less likely to contribute to future generations(5–7), given declining population survival due to starvation, disease, the elements, or predation. Modeling suggests that the fitness effect for humans of a phenotype-altering trait at 40 and 50 years of life are 1/5^th and 1/20^th respectively of the effect at age 20 (8), whether this fitness impact is positive or negative. In fact, for most of our evolutionary history, less than ~5% of humans lived past ~50 years of age (9,10). Moreover, George C. Williams proposed that inherited mutations with early life benefits but late life costs contribute to aging, in a process referred to as antagonistic pleiotropy(11). Phenotypes that benefit reproductive success early in life, but promote aging phenotypes at older ages, have likely been under positive selection throughout animal evolution. While these concepts may seem archaic, they are the foundation upon which our understanding of aging and aging-associated diseases (like cancer) must be framed.

Natural selection does not act to prevent all somatic evolution or to keep us disease-free beyond what is necessary to maximize our fitness(3). Considering tumor suppression, the strength of selection for mechanisms to prevent mutation-driven clonal expansions and malignant evolution will be proportional to the negative fitness effects (on the organism) of such somatic evolution (Figure 1). As such, the fitness cost of malignant evolution would clearly be greater at young ages than at older ages (Box 1). We also must consider the costs of tumor suppressive mechanisms, as the greater the cost (e.g. energetic) the more such a mechanism will be used sparingly. The ability of natural selection to assess the cost-benefit ratio of every trait (within the limits imposed by drift) has been essential for the evolution of our ability to mostly (but not entirely) avoid death or debilitation from cancer through years where reproduction was traditionally most likely. We will also discuss tradeoffs, such as traits that can increase cancer risk (e.g. inflammation) but are maintained due to their fitness benefits (e.g. tissue repair and combating infections).

Figure 1. Maintaining the soma to minimize the fitness impact of disease.

The relationship between somatic investment and somatic decline (with increased risks for cancers and other diseases) has been sculpted by evolution based on historic rates of survival and consequent reproductive success. Survival rates for humans are based on fossil data(9), reflecting high early life mortality rates followed by steady declines in survival due to extrinsic hazards and resource limitations. The curve for somatic investment is based on evolutionary theory, and roughly represents the strength of biological mechanisms acting as barriers to cancer (and other age-related disease). The cancer curve is based on SEER data for cancer incidence, but similar kinetics are observed for SEER data on cardiovascular disease (albeit shifted to the right by a few years).

As we will further discuss, such a cost-benefit assessment is at play every day in our bodies, with new traits resulting from genetic and epigenetic changes influencing the fate of the affected cell and its progeny (a somatic clone) in a context-dependent manner. The rules of such engagement have been established over millions of years of animal and mammalian evolution. As such, the evolution of strategies for disease avoidance and somatic maintenance, the waning of which is associated with aging, varies across animal species, in line with their evolved life histories. Mammals where extrinsic hazards are high (e.g. mice and rabbits) take a “breed early, breed often” strategy and thus exhibit aging (functional decline) earlier under protected conditions (such as in our labs), noting that survival to such ages would be quite rare in the wild. Mammals with lower extrinsic hazards (e.g. humans, elephants, bats and naked mole rats), using intelligence, size, flight, or underground burrows to avoid predation, have accordingly evolved much longer potential lifespans, as lower extrinsic mortality afforded opportunities for extended reproductive success (recognizing that other factors certainly contribute to the evolution of lifespans). Understanding both the common and group-specific cancer avoidance strategies from across the animal kingdom could illuminate novel approaches to lessen the impact of cancers on us(12). Cancer risk and tissue decline increase at older ages across species(13–16), and can be thought to scale with potential lifespan, being largely relegated to periods when contributing to future generations was less likely. For example, mice rarely survive past 1 year in the wild(17), and most cancers develop in mice well past a year old(18), including for genetically diverse mice from the collaborative cross(19). Thus, the age-dependence of cancers in mice scales across a ~2-year period relatively proportionally to the risks for humans across ~80-year period. Finally, the rates of cancer and other diseases of aging differ between men and women, often showing higher and earlier occurrence in males(20,21), likely relating to fitness tradeoffs in males (favoring aggression, risk taking and greater muscle mass in males, despite higher associated mortality), although modern differences in lifestyles and exposures (e.g. smoking, alcohol intake, and diet) also contribute.

The extreme inefficiency of cancer – evolved mechanisms of tumor suppression.

While the focus of cancer biologists is typically on understanding why we develop cancers (and how to prevent them), we should also marvel at how we can develop into a body with 5–7 trillion nucleated cells, with each cell in our body, on average, accumulating about 20 mutations per year(22), while manifesting an impressively low rate of cancer development for about half a century. How are we so good at not getting cancer?(23) This feat is particularly astounding when we consider that each of us will accumulate at least 100 billion cells with known cancer-associated mutations in our histologically normal tissues by the time we are in our 60’s(24), the result of ubiquitous clonal expansions. We and other animals have accomplished this impressive tumor suppression, which is strongest through years of likely reproduction, via the evolution of barriers that act at many levels, from avoiding mutations, preventing their persistence, and limiting their clonal expansion to disfavoring metastatic colonization. Extending from the earlier theoretical work of Gatenby and Gillies(25), tumor suppressive barriers to be discussed include – 1) genome and epigenome maintenance, 2) regulated cell death, 3) senescence and telomere attrition, 4) cell competition and stabilizing selection, 5) immune surveillance and control of inflammation and 6) metastatic suppression and dormancy (Figure 2). We will discuss how most of these barriers change at older ages, while other barriers appear relatively constant throughout life. We should note that these barrier mechanisms will be imperfect, either due to biophysical limits (error is inevitable), the cost of further improvement (e.g. energetic costs that might be better invested elsewhere, such as for fecundity), and the limited resolution of natural selection due to the drift barrier (e.g. for mutation rates(26)).

Figure 2. Barriers to cancer progression delay most cancer risk till older ages.

Barriers to cancer progression are depicted, recognizing that different barriers will exhibit different efficiencies, and even depend on context – cell type, (epi)mutations, and tissue environment. As a general rule, barrier function tends to decrease with aging, but exceptions exist (such as telomere attrition) and we lack insight into the age-dependence of other barriers (like for oncogene-induced senescence). While barriers are depicted at discreet points in cancer progression, some of these may act at multiple points, like “control of inflammation”. Natural selection has acted to erect these barriers to minimize the negative fitness costs of malignant growth, leading to the late-life predominance of cancers. A combination of compromised barriers with age and altered cell intrinsic and cell extrinsic biological processes (red arrow) may facilitate the bypass of these barriers and allow greater numbers of clones to progress to more advanced stages (thicker blue arrows) as well as promoting larger more aggressive and more invasive tumors in old age. Created in https://BioRender.com

Finally, epidemiological data suggest that cancer incidence decreases at very old ages (past 80 years old, Figure 1). There are some possible explanations for this phenomenon including competing causes of mortality (e.g. cardiovascular diseases) and altered biological processes (e.g. telomere attrition, cellular senescence and stem cell exhaustion)(27). Still, given that the reduction is predominantly observed in cancers requiring invasive diagnostic tests, and that autopsy observations do not corroborate this reduction, decreased cancer incidence at older ages may simply be due to under diagnosis(28). Regardless, here we focus on the general phenomenon of increased cancer incidence at older ages and its root causes.

(Epi)Genome maintenance: Limiting somatic mutations and epigenetic change

Natural selection requires variability in heritable phenotypes to exert change in a population, whether for organisms or cells within somatic tissues. For the latter, mutations and epigenetic alterations, referred to in aggregate as (epi)mutations, can lead to heritable phenotypic change. While most (epi)mutations will be neutral, those that alter phenotype within a replicating cell population will be subjected to selection, either negative (purifying) or positive. Far more (epi)mutations will reduce cell fitness than increase it. As will be discussed below, animals have evolved mechanisms to disfavor the persistence of cell clones with mutations that alter phenotype, both to maintain tissue function and to limit oncogenesis. In this section, we will discuss how effective maintenance of the genome/epigenome can limit the generation of heritable phenotypic diversity to maximize fitness and minimize cancers in youth, and how these mechanisms might wane at older ages.

Recent studies reveal how DNA mutation load (at least for single nucleotide variants, or SNV) accumulate in human and other mammalian tissues linearly with age, with the slope (the rate) similar at young and old ages(22,29–32). Contrary to expectations, these studies show that most mutations accumulate independently of cell replication, as similar mutation frequencies are observed in tissues with very different turnover rates (from the colonic epithelium to neurons). Historically, the increasing risk of cancer at older ages has been modeled through the requirement for the sequential acquisition of multiple events (thought to be (epi)mutations) that stimulate clonal expansion – the multi-stage model of carcinogenesis(33). While the requirement for multiple (epi)mutations for cancer development should indeed contribute to the relegation of most cancers to older ages, additional mechanisms of tumor suppression to be discussed here are clearly playing major roles.

Studies using single cell whole genome sequencing reveal copy number alterations (CNAs) in normal human breast tissue (in ~3% of cells from every analyzed sample, and largely in luminal cells), with specific CNAs recurrently observed across donors indicating positive selection and convergent evolution(34,35). The frequency of these CNAs increases with age(35), consistent with observations of more structural rearrangements at older ages in mice(36). Similar chromosomal alterations, such as gain of 1q and loss of 16q, are present in breast cancers. While most SNVs will not impact cell fitness, and thus the total SNV count in a genome can be used to determine rates of mutation accumulation, the same cannot be said for CNAs and other structural rearrangements. We thus need to consider how possible changes in selection at older ages (such as reduced purifying selection or increased positive selection) might contribute to alterations in the frequency of structural rearrangements in the genome. Regardless, the frequency of cancer-associated chromosomal alterations in normal tissues far outpaces the incidences of the cancers that arise from these tissues, reflecting the myriad of tumor suppressive hurdles that limit the progression of these aberrant clones to cancers.

A key question in evolutionary cancer biology is why cancers are not more common in larger animals (more cells provide more opportunities for cancer evolution) or longer-lived animals (more time for such evolution), known as Peto’s Paradox(37,38). A recent study used whole genome sequencing of intestinal crypts from a variety of mammals to determine mutation accumulation across the lifespan(30). As intestinal crypts are relatively clonal (due to drift kinetics within very small stem cell pools for each crypt), detected mutations largely reflect those present in a single stem cell (albeit in the recent past). Strikingly, different mammals accumulate a similar number of mutations within a lifespan (within a few fold), despite a ~100-fold variance in lifespan. Interestingly, mutation accumulation did not substantially scale with body size (which varies 40,000-fold across these mammals), beyond that contributed by lifespan. Previous studies using sequencing of single cell-derived organoids from both human and mouse tissues (liver, small intestines and large intestines) showed similar numbers of mutations accumulating within a lifespan for mice and humans, noting that mice still accumulate up to 5-fold fewer mutations (albeit with a 30–40-fold shorter lifespan)(29,32). Mouse hematopoietic stem cells (HSC) have roughly 3-times the mutation rate per year than human HSC, which equates to about a 10-fold reduction in total mutation accumulation per cell in mice as compared to humans at the end of life (~150 vs ~1500)(39). While speculative, natural selection has perhaps acted to limit mutation accumulation to less than a few thousand mutations per somatic cell genome in a lifetime, noting that selection for limited germline mutation rates (affecting the fitness of progeny) may also play a role (as discussed in (30)). DNA repair efficacy has been shown to scale with lifespan(40), which could contribute to establishing this ceiling. Given that this ceiling has been calibrated for lifespan, not body size, natural selection may have acted to limit mutation accumulation to maximally maintain tissue functionality (i.e. to prevent aging). While limiting mutation accumulation should also reduce cancer risk (by reducing the generation of somatic heritable variability), other mechanisms must also act to reduce cancer evolution in larger animals given that greater cellularity should provide more potential cancer initiating events. Within species there is evidence that the efficiency of DNA repair is compromised with aging, although the data are mostly derived from in vitro experiments or based on indirect measures ex vivo(41,42), conflicting with data discussed above that mutation (SNV) accumulation is quite linear across the lifespan. Still, it should be recognized that DNA damage can persist for years in stem cells(43) and can cause age-related phenotypes independent of whether it is converted into a DNA alteration (discussed below)(44,45).

The rate of epigenetic disorder also scales with maximum lifespan across mammals(46). Epigenetic drift is speculated to reduce stem cell and tissue functionality(47), which should increase tumorigenesis such as by promoting the somatic evolution of adaptive phenotypes (some of which could be malignant), altering immunity (less adaptive, more inflammatory), and reducing the somatic fitness of stem cell pools. Epigenetic patterns across tissues degrade across a lifespan, including in stem cells(48), and indeed such changing patterns are tightly linked with chronological age in humans(49). Unlike DNA mutations, changes in DNA methylation at CpG sites are highly associated with organismal growth, with a much more rapid rate of change during ontogeny followed by a slower linear period of change post-maturity, with no obvious change in this rate at older ages(50). Interestingly, one’s deviation from these patterns (for DNA methylation at selected CpG sites) can indicate decelerated or accelerated aging, providing a “biological age” that is predictive of disease risks and overall survival(51). In general, DNA methylation tends to decrease genome-wide with age, but with increases at CpG islands in genes. Some CpG sites show increased variability in methylation with age. There are also links between DNA damage and epigenetic deregulation with age. Stochastic endogenous DNA damage with age leads to RNA polymerase stalling, promoting aging-associated gene expression patterns and likely aging phenotypes(44). Experimentally induced double stranded DNA breaks can also promote epigenetic deregulation and aging-like phenotypes(45) (recognizing complications with the model(52)). In turn, changes in epigenetic marks can result in reduced DNA repair(53,54).

Histone modifications also exhibit alterations with age, although these alterations vary in directionality for different modifications and across different organisms and tissues(53,54). Aging is not uniform across individuals, tissues or cells(55). Importantly, old cells show greater epigenetic heterogeneity than young cells, whether measured by DNA methylation patterns or gene expression at the single cell level(53). Single cell analyses of gene expression in the lungs(56), in HSC pools(57), and in various blood cells(58) serve as representative examples of increased epigenetic heterogeneity at older ages. Studies of monozygotic twins demonstrate how patterns of DNA methylation are strikingly similar at birth but extensively diverge with age(59). The importance of the epigenome in aging is emphasized by the ability of epigenetic reprogramming to reset somatic cells to a more youthful, pluripotent phenotype(60), including for senescent cells from very old humans(61).

One consequence of altered DNA methylation at older ages, histone modifications, and heterochromatin formation is the awakening of endogenous retrotransposons (ERT; largely LINE1/L1 elements in humans) in the genome, leading to cytoplasmic RNA and DNA. Cytoplasmic RNA can activate MDA5, RIG-I and TLR3 sensing pathways and cytoplasmic DNA can activate the MDA5 and cGAS/STING pathway, each contributing to a type I interferon response and aging-associated inflammation(62,63). Retrotransposition can also induce DNA damage, which can elicit inflammatory responses and even promote further ERT awakening. As we will discuss below, induced “inflammaging” can alter cancer evolutionary trajectories.

Increases in (epi)genetic heterogeneity across cells in tissues as we age should engender more cell heritable variability upon which selection can act, providing one key ingredient for somatic evolution that can lead to malignancies (Figure 2). Genetic and epigenetic changes in somatic cells will be subjected to neutral, positive, or negative (purifying) selection, with positive selection being the rarest of fates for these (epi)mutations (although there is a paucity of data measuring negative selection in the soma). As will be discussed below, selection is highly dependent on context, with substantial influence from the state of the microenvironment and the age of the host. Whether the ability of animals to limit (epi)mutation occurrence throughout life is constant or reduced in post-reproductive periods, additional checks on the ability of these mutated cells to persist, expand or further evolve do appear to become weaker at older ages.

Regulated cell death: Elimination of damaged and potentially pre-malignant cells

Cells can be eliminated by means of accidental death, when faced with irreparable damage, or through of one of the regulated cell death (RCD) pathways – evolutionary conserved molecular programs that lead to the removal of the cell when it serves a beneficial purpose for the organism. The most well described RCD pathway is apoptosis, a caspase-dependent process initiated either by mitochondrial outer membrane depolarization and release of cytochrome C or by activation of “death receptors” such as FAS and TRAIL by extracellular ligands. However, many other mechanisms of RCD have been described(64,65). Although a comprehensive examination of the role of individual RCD in cancer and aging goes beyond the purpose of this review, extensive evidence suggests that different RCD pathways can play a tumor suppressor role(66,67), thus justifying the general view that RCD constitute one of the barriers to cancer development. As the most studied RCD, there is abundant evidence for a role of apoptosis in cancer suppression, substantiated by the fact that a fundamental aspect of cancer evolution involves suppression of apoptotic regulators by multiple means(68,69) (Figure 2).

The cancer suppressive role of RCD is also evident by the fact that the expression of activated oncogenes in non-transformed cells often results in induction of cell death signals(70). In particular, it is well established that increased MYC activity can result in promotion of apoptosis through multiple mechanisms(71). Furthermore, the inactivation of the retinoblastoma protein (RB1) in non-physiological contexts also results in apoptotic induction in fibroblasts mediated by active E2F1(72). Phosphorylated RB1 protein can directly mediate intrinsic apoptosis at the mitochondria by interacting with BAX(73). Interestingly, in contrast to RB1 and MYC, active HRAS induces a different type of cell death, either a non-well-defined pathway involving pinocytosis(74) or autophagic cell death(75). It should be noted that these pathways have been observed in human and murine cell lines in vitro. While a direct observation of oncogene-induced cell death in vivo is currently lacking, the genetic disruption of RCD pathways that frequently accompanies oncogenic transformation in human cancer argues for the relevance of these cancer-suppressing mechanisms(76).

An important question is whether deregulated RCD pathways in aging might play a role in the increased incidence of cancer in the elderly population. Multiple lines of evidence suggest that different RCD pathways are indeed altered during aging. Apoptosis sensitivity is generally decreased in aging suggesting an age-dependent failure of apoptosis barrier function(77) (Figure 2). An elegant study has shown that multiple tissues in mice and humans, such as brain, heart and kidney, possess mitochondria primed for apoptosis in early life (by BH3 profiling), which become more refractory to apoptotic cues in adulthood, although elderly individuals were not examined(78). This study may suggest the existence of stringent quality control during development that wanes in adulthood. While this decrease of apoptosis as a barrier temporally precedes the increase in cancer incidence, we should not discount the possibility that malignant clones develop in adulthood and when not cleared by RCD might be the founder of frank cancer manifesting in old age. Additional studies have also shown that apoptosis after irradiation was decreased in the spleen but not the colon of old mice when compared to young mice(79). In human serum from individuals 20 to 80 years old, aging is correlated with decreased positive apoptosis regulators and increased negative regulators, and this trend was exacerbated in patients with breast cancer and prostate cancer(80). Indirect evidence exists that apoptosis markers are reduced in bone marrow mesenchymal stem cells(81) and in adipose tissue-derived mesenchymal stem cells of old mice(82). This suggests that the function of apoptosis as a cancer barrier may weaken with age. Regarding other RCD pathways, there is some evidence that necroptosis and ferroptosis might increase at older ages(83,84). In contrast to apoptosis, which is generally not or mildly immunogenic, necroptosis and ferroptosis are associated with release of damage-associated molecular patterns (DAMP), which can play either a promoting or suppressing role on cancer depending on context through contributions to age associated-inflammation and feedback modulation of additional RCD pathways (83–86). Given that most cancer targeting approaches rely on some form of RCD to eliminate cancer cells, it is important to understand how aged cancer and aged tissues respond to specific RCD-inducing treatments and how different regiments would be most beneficial in younger vs older patients.

Senescence and telomere attrition: Preventing the outgrowth of cells with activated oncogenes or eroded telomeres

Cell senescence is a permanent state of cell cycle arrest that can be induced by specific stimuli including oncogene activation and DNA damage and is associated with multiple cellular alterations, most notably the acquisition of the senescence-associated secretory phenotype (SASP): the increased production and secretion of multiple inflammatory cytokines, chemokines, proteases as well as extra-cellular vesicles, which can result in important microenvironmental alterations(87,88). Senescence has been shown to be a prominent initial barrier to oncogenic transformation(89,90). In a well characterized model of liver carcinogenesis, mutant RAS-induced tumor formation is impaired by senescence induction mediated by p53 activation, which results in a SASP phenotype that recruits both innate and adaptive immune cells, leading to the clearance of senescent cancer cells(91–94).

On the other hand, multiple studies in animal models have shown that senescent cells have a causal role in promoting various aspects of aging and aging-associated diseases. Induction of senescence or transplantation of in vitro generated senescent cells can promote aging phenotypes, while clearing senescent cells with drugs (senolytics) or other genetic approaches to eliminate senescent cells restores a youthful phenotype in progeroid or aging mice, and even delays cancer development and prolongs lifespan(95–99). Cells expressing senescence markers have been shown to accumulate with aging in various human tissues(100,101). Notably, senescence induced by a variety of stimuli, including oncogene induction and infection with SARS-CoV-2, can spread to nearby cells by means of SASP factors creating a positive feedback loop(102–105). Interestingly, aging-associated expression of endogenous retroviruses (ERV) can induce cellular senescence and worsen aging phenotypes in mice in a cGAS-STING dependent mechanism(106). The eventual elimination of senescent mutant cells and even small tumors by the immune system is important to prevent the acquisition of a more malignant phenotype(107,108), a concept supported by the observation than premalignant tumors show more frequent expression of senescence markers than advanced cancers(109). It remains to be determined whether the accumulation of senescent cells is the result of higher rates of senescence induction (by accumulation of DNA mutations or telomere attrition, see below) or impaired clearance of senescent cells in aged individuals (or both). The immune system plays an important role in clearing senescent cells (110,111) and age-dependent dysfunction in immune cell function may promote the accumulation of senescent cells with aging(112–114).

A large body of literature supports the fact that senescent cells and in particular the SASP can promote cancer in various ways. Initial experiments have shown that senescent fibroblasts act in a paracrine fashion to promote the growth of cancer cells lines in co-culture experiments and tumor formation in xenografts(115). Oncogenic mutations such as in HRAS and TP53 can produce an amplified SASP from senescent cells which potentiate the induction of epithelial-mesenchymal transition in non-aggressive breast cancer cell lines(116). Several specific SASP factors have been implicated in the pro-tumorigenesis effects in various contexts, including amphiregulin (AREG)(117), growth differentiation factor 15 (GDF15)(118), matrix metalloproteinases (MMP) and hepatocyte growth factor (HGF)(119), interleukin-6 (IL-6)(120), and ephrin type-A receptor 2 (EphA2)(121) (as a few specific examples), although most of these studies were conducted in young mouse models. Understanding how specific SASP factors are altered in the microenvironment of aged tissues, particularly for humans with or without cancer, will be critical to guide more effective prevention and treatment strategies. Interestingly, it has been shown that serum from elderly donors can promote the proliferation of breast cancer cell lines in vitro in an IL-6 dependent mechanism, although it was not shown if the source of IL-6 was senescent cells(122). Clearly, more studies of the interaction between senescence (and the SASP) and the evolution of cancers in aged contexts will be critical for untangling cause-and-effect relationships.

SASP factors have also been shown to promote immunosuppression – in the skin, IL-6 expressed from senescence stromal cells alters the immune system to prevent the clearance of squamous cell carcinomas(123). Notably, such an immunosuppressive microenvironment was also observed in the skin of aged cancer-free individuals. In a colorectal cancer model, immune suppression was due to high expression of CXC motif chemokine ligand 12 (CXCL12) and colony-stimulating factor 1 (CSF1), which prevented the infiltration of cytotoxic T-lymphocytes and led to their exhaustion(124). In lung cancer, tumor growth factor β (TGFβ) acted in concert with a hypoxic microenvironment to generate a 14-factor immunosuppressive SASP which reduced the effect of immune checkpoint activators(125). Similarly, the clearance of senescent cells with ABT263 restored the effectiveness of anti-PD1 treatment in mice injected with colorectal cancer cells or lymphoma cells(126). While it is apparent that the immunosuppressive effect of senescence in different contexts is dependent on different flavors of the SASP, in all aforementioned cases the end result was an increase in infiltration of myeloid-derived suppressor cells (MDSC) expressing CD11b and Ly6G and a reduction in CD8⁺ T-cell numbers associated with reduced expression of activation markers and increases in markers of exhaustion, which is similar to the immune changes observed in aging (discussed below).

Telomere erosion can lead to senescence, playing a critical tumor suppressive role (Figure 2). Telomeres are the protective DNA elements at the ends of chromosomes. Due to the “end replication” problem, at every cell division human telomeres shorten by 100–200 bp. Cultured human cells not expressing telomerase can replicate about 50 times before undergoing senescence(127,128). Notably, senescence is induced when telomeres are critically short but not exhausted; if senescence is bypassed, cells reach the “crisis” point where eroded telomeres induce chromosomal end-to-end fusion and recombination, at which point a second roadblock is cell elimination through autophagic cell death(129) and concomitant innate immune activation(130). It has been clear for many years that the Hayflick limit of cultured cells is not directly applicable to somatic cells in vivo, which are estimated to have a much larger number of replications throughout life(131). Indeed, many stem and progenitor cells express telomerase to maintain lifelong self-renewal.

Telomerase activity, however, is strictly controlled, and it is quickly turned off after differentiation to limit the number of replications each clone can undergo(132), with clear implications for cancer suppression. The association of telomere length and cancer risk is somewhat controversial, mostly due to the technical complication of accurately measuring telomere length at clonal levels. However, a large Mendelian randomization study showed that small nucleotide polymorphisms associated with longer telomeres are associated with increased risk of various cancers(133). Another Mendelian randomization study showed that increased telomere length is a risk factor for clonal hematopoiesis of indeterminate potential (CHIP), a pre-malignant expansion of mutant hematopoietic cells(134).

The necessity for malignant clones to resolve the telomere erosion problem is also clear from the fact that the vast majority of human cancers express telomerase, often due to point mutations in the telomerase promoter. In other cases, telomeres are extended by the “alternative lengthening of telomeres” (ALT) process which exploits the homologous recombination machinery(132,135). Unlike humans, small rodents like mice have much longer telomeres and less restricted telomerase regulation. In fact, lethal tumors can form in mice without telomerase expression, in contrast to humans(136,137), suggesting that regulated telomerase expression in large but not small mammals is a protective mechanism against neoplastic transformation. Indeed, across different mammalian species, telomerase expression becomes increasingly restricted proportional to body mass (but not lifespan)(138). The role of aging in affecting telomere length and cancer protection is still unclear. Telomeres show a trend toward reduction with age, but the correlation is weak and most pronounced during development(31,139), suggesting that the rate of cell expansion rather than chronological age plays a larger role in telomere shortening. This observation also suggests that the effectiveness of telomere attrition as an anticancer mechanism should not be reduced in old age (Figure 2); whether telomere shortening results in a stronger barrier function at older ages has not been tested. A different perspective to consider is that the progressive accumulation of senescent cells with critically short telomeres with age might compromise the tissue microenvironment and promote the clonal selection of cells with longer telomeres and/or reactivated telomerase, leading to increased chances of malignant transformation. This scenario can potentially explain why patients with telomere-shortening syndromes show an increased risk of various cancers(140). Thus, it seems that telomere length has ambivalent roles in cancer, with too short telomeres promoting senescence and aging phenotypes in normal tissues and too long telomeres facilitating the expansion and potential transformation of mutant clones. This dynamic needs to be carefully considered when evaluating the potential impact of telomerase-targeting treatments in cancer.

Cell competition and stabilizing selection: Maintenance of youthful tissue landscapes, limiting oncogenesis

Our tissues are under strict homeostatic control, with defined numbers of each cell type cooperating to maintain tissue structure and function. In addition, as will be argued here, the maintenance of this normal structure in youth should be tumor suppressive (and generally disease suppressive). A key mechanism to maintain tissue structure and function is through cell competition between homotypic cells(141), where cells with higher relative fitness (winners) can actively eliminate cells with lower fitness (losers)(142,143). Mutations that impair or promote cell competition mechanisms have been shown to reduce or increase, respectively, lifespan in Drosophila melanogaster(144). Winner cells can engulf and digest loser cells, or force loser cells from the self-renewing monolayer. For example, a cell can be recognized as “less fit” should it express less Myc, as studied extensively in D. melanogaster(145–147). Protein translation efficiency is also a key metric of cell fitness, from flies to mice(148). In mice, HSC clones with either increased or decreased translation are clonally eliminated(149). Thus, maximal HSC fitness requires a narrow range of translation. Given that many oncogenic events promote translation(150), this mechanism not only eliminates poorly functioning HSC but also those with potentially oncogenic mutations. In fact, cell competition has been shown to eliminate cells with potentially oncogenic mutations in multiple organisms(151,152), including in mammalian epithelial tissues(153) (Figure 2).

Tissue homeostatic mechanisms can further eliminate small tumors. Studies in mouse models demonstrate how tumors are eliminated through competition with mutant clones in the phenotypically normal esophageal epithelium(154). Indeed, increasing the competitiveness of competing clones by activation of the Notch pathway increases the elimination of nascent tumors. Thus, clonal expansions, such as those driven by mutations inactivating the Notch pathway, can have tumor suppressive roles. Additional studies have shown how early growths with mutations activating Wnt/β-catenin or KRAS in the skin can be actively extruded, in a process requiring wildtype neighbors(155). Interestingly, recognition of deformed tissue architecture appears sufficient for eliciting elimination, attesting to the power of a tissue (the skin) to correct aberrant structures to maintain function and limit malignancy. Recent studies have also shown that pharmacological interventions such as metformin can alter the dynamics of cell competition in PI3KCA mutant esophageal cells or β-catenin and HRAS mutant epidermal cells in mice(156,157).

Critically, there is evidence that the efficacy of cell competition declines with age, due to both microenvironmental changes (whereby the “evolved” type is no longer the most fit) and due to intrinsic decline in stem and progenitor cell fitness(3). Studies in the murine hematopoietic system have shown that young hematopoietic stem and progenitor cells (HSPC) can outcompete old HSPC in the reconstitution of hematopoiesis(158). Importantly, while oncogenic mutations (such as in Nras or Bcr-Abl) are selected for in an aged hematopoietic system, providing young competitor HSPC can prevent the expansion of these oncogenically-initiated clones(159,160). We have proposed that healthy and youthful tissues actively select against change in heritable phenotype – essentially, stabilizing selection at the somatic level(23,161). In contrast, aged or otherwise damaged (e.g. through smoking) tissues present a very different adaptive landscape, promoting selection for new phenotypes that are adaptive to this altered tissue environment. We propose that the aging-dependence of “adaptive oncogenesis” represents one of multiple mechanisms whereby cancer risk is largely delayed till older ages.

Given that maintaining healthy tissue, in addition to maximizing functionality of the tissue, can also prevent the establishment and expansion of malignant clones, understanding the cell competition pathways that maintain tissue integrity is paramount. In the skin, cell competition mechanisms can prevent the persistence of cells and clones with DNA damage, which induces downregulation of collagen COL17A1 (an integral component of hemidesmosomes), leading to poor attachment to the basement membrane (required for symmetrical stem cell divisions needed for self-renewal)(162). Thus, COL17A1 serves as a sensor of DNA damage (and likely other damage or stress) that promotes loss of the damaged stem cell through differentiation. Clones resulting from cells that received DNA damage fail to be maintained in the stem cell pool, instead contributing to more differentiated layers of epidermis (and doomed to be shed). This mechanism can help maintain the competitiveness of the epidermis, but declines at older ages. While speculative at present, this mechanism should also promote tumor suppression as more fit stem cells should also aid in the elimination of stem cells with potentially oncogenic mutations. One key mechanism to maintain tissues is autophagy, and modulation of autophagy has been shown to control lifespan across animals(163). Boosting autophagy in mice, whether through mutation of Becn1 (encoding Beclin-1)(164) or overexpression of Atg5(165), has been shown to reduce aging-associated cancer incidence, although the link with improved tissue maintenance has not been established.

Strikingly, studies in the last few years have shown how as we age our tissues become dominated by clones driven by known cancer-associated mutations(1,24,166). Such clones have been characterized in various tissues, including the hematopoietic system, liver, lung, bladder, skin, colon, and endometrium, largely in the absence of overt phenotypic manifestations. That older individuals can possess over 100 billion cells with known cancer-associated mutations(24) is both a testament to the power of tumor suppressive mechanisms to prevent the malignant progression of the vast majority of these cells and a demonstration of the loosening of mechanisms to limit these expansions at older ages. It is critical that we better understand the risks that these clonal expansions confer, both in terms of cancers and other diseases, and how such risks can vary dependent on the associated mutations, clonal size, and other features. Given the wide availability of blood samples with genomic or exomic sequencing, studies of clonal hematopoiesis (CH) are the most advanced in this regard. Expansion of CH clones is largely restricted to the elderly, with CHIP (where variant allele frequency for an associated mutation exceeds 0.02) most prevalent in individuals over 70(167,168). While CHIP is often characterized by mutations in known leukemia driver genes, CHIP is associated with various maladies of old age, including cardiovascular disease, cancers in general (in addition to leukemias), diseases of the liver and lung, frailty and overall mortality(168–170). Whether the same is true for solid tissues awaits further research, although such associations will be more difficult given the much smaller sample sizes (at least for analyses thus far).

Clonal expansions are not only associated with SNVs and other small mutations like indels. A pan-tissue analysis using RNA-seq data from the Genotype-Tissue Expression (GTEx) project revealed an age-dependent increase in clonally-expanded chromosomal alterations across multiple tissues; 25.7% of individuals had a detectable alterations (>7–10% cell fraction) in at least one tissue(171). Chromosomal alterations are also well characterized in CHIP(172). Single cell whole genome sequencing of normal human breast tissue reveals that some CNAs are associated with clonal expansions with age, and cells with these CNAs localize to ductal and lobular structures with normal histopathology(35). In one case, the mammary epithelial clone with 1q gain/16q loss, through an unbalanced translocation (der(1;16)) apparently acquired in adolescence, expanded to occupy a large fraction of the tissue before subclonal evolution led to a breast cancer(173).

A key question is whether clonal selection for cancer-associated mutations is constant throughout life or differential. Some reports have argued that a constant rate of selection on driver mutations throughout life can explain the increased clonality of hematopoiesis (such as CHIP) at older ages(31,174,175). On the other hand, there is evidence from mouse models for differential selection for multiple mutations in HSPC, with stronger positive selection in older ages(159,160,176–179), in some cases shown to be due to mutation-conferred resistance to an inflammatory microenvironment. Importantly, we speculate that the pattern of selection could be different for different mutations in different genes, as natural selection will have acted to sculpt tissue landscapes to disfavor somatic changes to the extent that maximizes organismal fitness. Evidence for this idea can be found in patterns of clonal expansions in normal tissues, with (for example) mutations in NOTCH1 in the esophagus and DNMT3A in the blood arising much earlier in life than mutations in TP53 in the esophagus and SF3B1 in the blood, coinciding with the much greater risk of malignancies in those tissues conferred by the latter mutations(180–183). Notably, risk for cardiovascular disease and all-cause mortality is associated with the same CHIP mutation (and other) parameters associated with risk of hematopoietic malignancies (184,185). Thus, while speculative, mutations with greater potential to impair organismal fitness (e.g. in TP53 and SF3B1) appear to exhibit stronger negative selection in youth, perhaps followed by positive selection (“unleashed” from the pressures of natural selection) at older ages, consistent with mathematical modeling(186).

Characterizing the dynamics of this aging-dependent selection is valuable not only to understand relative cancer risk at different ages but also to discover biological processes that can be manipulated to reduce the chances of cancer formation or progression. Notably, cell competition in cancer is an area that is currently under-investigated and no current treatment strategies to our knowledge exploit this potent cancer suppressive mechanism. Finally, it is important to note that this review should not be considered a comprehensive catalog of all barriers to cancer evolution, as there are certainly additional barriers yet to be discovered or that are insufficiently characterized in the context of aging (such for metabolic changes, see Box 2).

Box 2. Aging-dependent metabolic changes and cancer pathogenesis.

Aging is associated in several tissues with metabolic changes such as a global reduction in mitochondrial oxidative phosphorylation (OxPhos) and accumulation of metabolites like 2-hydroxyglutarate, fumarate and succinate which have been suggested to play a role in epigenetic dysregulation by modulating the role of several chromatin modifying enzymes(187). These changes may depend on age-associated reduction of intracellular NAD+ level and reduced mitochondrial biogenesis through increased HIF1α signaling(188) or a PDK4-dependent aerobic glycolysis phenotype in senescent cells(189). Aging is associated with a pseudo-hypoxic state that shifts metabolic flux from mitochondrial OxPhos to glycolysis, a phenotype proposed to provide a metabolic advantage to cancer cells by delivering lactate and altering the microenvironment(188,190). In melanoma, a shift to reduced OxPhos is associated with increased metastatic burden, particularly for the liver(191) An important age-associated metabolite with oncogenic properties is methylmalonic acid. Methylmalonic acid is increased in the serum of elderly subjects and has been shown to promote a more aggressive phenotype and drug resistance in breast and lung cancer cells(192) and to promote epithelial to mesenchymal transition in colorectal cancer(193), in lung cancer and melanoma(194). These results further suggest that alteration of the metabolic state of cells and tissue in aging might represent an erosion of an additional barrier to tumorigenesis.

Immune surveillance and control of inflammation

A primary and ancestral role of the immune system is in host defense from pathogens and wound healing. Maintenance of a healthy immune system in youth also contributes to tumor suppression. Both inflammation and wound healing are good examples of antagonistic pleiotropy, serving essential functions throughout life while some of the same mediators can contribute to tumor progression at older ages (often in chronic form)(195).There is extensive evidence that the immune system has profound effects on the emergence and progression of cancers of different kinds. In general terms, components of the innate immune system such as macrophages and neutrophils have ambivalent roles given that they can acquire pro- or anti-inflammatory specifications that can promote or prevent the expansion of cancer cells depending on their cytokine milieu and their regulation of the microenvironment. Cells of the adaptive immune system, prominently natural killer (NK) cells and CD8⁺ T cells, generally elicit strong cancer suppressive functions but exceptions exist(196).

Mice and humans with compromised immune systems are at increased risk for certain cancers(197). Eliminating B- and T-cells in mice, such as through disruption of the Recombinase Activating Genes (Rag1 and Rag2) necessary for rearrangement of immune receptor genes, increases the rates of both spontaneous and carcinogen-induced cancers(198). Additional studies in mouse models reveal how the adaptive immune system maintains occult cancers in an equilibrium state(199). Still, the fact that Rag1 or Rag2 knockout mice do not develop even more cancers suggests that adaptive immunity is just one of multiple barriers to cancer development. Indeed, mice engineered to be NK cell deficient exhibit a substantially increased risk of cancer development(200), and mice lacking γδ T-cells are highly susceptible to multiple regimens of cutaneous carcinogenesis(201). It has been observed that patients with various primary immune deficiencies display only a marginal increase in cancer incidence and only for certain cancers(202), noting the cell-specific nature of the studied immunodeficiencies (mostly affecting B-cells).

Immune surveillance is clearly directing the evolutionary trajectories of cancers. For example, different lung cancers evolve distinct mechanisms of immune evasion, from limiting immune infiltration to mutational loss or downregulation of histocompatibility loci (or B2M) to elimination of antigenic epitopes through copy number events(203). Additional studies show how lung adenocarcinoma pre-cancer evolution coincides with a shift from innate to adaptive immune to immune evasive phenotypes (Res Sq 2024.05.15:rs.3.rs-4396272). Moreover, individuals with a higher immune epitope burden (based on encoded immunogenic peptides and MHC Class 1 alleles) in their germline HER2 gene are less likely to develop HER2-positive breast cancer, but those that do develop a more aggressive and immune-cold cancer, suggesting an immune suppressive escape(204). Immunoediting and immune evasion are well described phenomena in mouse and human carcinogenesis(205). These observations and the clinical success of immune checkpoint inhibitors bolster the concept that the immune system presents a critical barrier to cancer evolution (Figure 2).

Aging-dependent changes in the composition and functionality of the immune system have been extensively characterized. In the bone marrow, HSC become progressively skewed towards myeloid differentiation to the detriment of lymphoid cell output in mice and humans(206,207). This effect seems to have a causal role in hematopoietic system aging since antibody-mediated clearance of myeloid-biased HSC restores young-like immune phenotypes(208). Furthermore, a high neutrophil to lymphocyte ratio is a negative prognostic marker in many solid tumors(209). With aging, thymic involution results in progressively lower output of naïve T-cells which result in a decreased T-cell receptor repertoire(210), exhibiting a good correlation with cancer incidence(211). Thymic involution is evolutionary conserved across many vertebrate species. It is not clear the reason why this phenomenon occurs but it has been speculated that the intensive energy requirement for continuous T-cell production and the risk of thymic cancer outweighs the benefits, especially considering that thymectomy in adult organisms do not significantly compromise overall health(212). Several other changes in T-cells with aging have been described including mitochondrial dysfunction, senescence and compromised effector plasticity (213). Interestingly, a population of granzyme-K-expressing T-cells expands with aging and shows transcriptional signature of terminally exhausted T-cells(214). In another study, an aged microenvironment drives the generation of a dysfunctional population of T-cells with impaired cytotoxic function(215), and remodeling of the collagen matrix in the aged skin has been shown to impede the migration of T cells(216). B-lymphopoiesis is also downregulated with age at multiple stages of development starting from HSC, further contributing to myeloid-biased hematopoiesis, and mature B-cells display significant alterations in subset frequency and function contributing to T-cell alterations and inflammaging(217). Aging also affects the function of NK cells(114) and dendritic cells(218). Still, there are only a few studies directly demonstrating how declining adaptive immunity in old age contributes to cancer pathogenesis.

As described in the previous section, multiple studies describe how senescent cells and their SASP can engender an immunosuppressive microenvironment that enables cancer progression. In addition, aging-associated chronic low-grade inflammation (also known as inflammaging) can have detrimental effects on the immune system leading to increased morbidity and mortality, including from cancer(219). Among several causes of inflammaging, the gut microbiota plays an important role, and the dysbiosis observed with aging shows a strong connection with immune system defects in old age(220). CHIP that is prevalent in older individuals can further promote inflammation, and some CHIP mutations (such as in TET2) have been shown to be selected for in HSPC by inflammatory conditions(178,221,222), leading to a feedforward loop likely contributing to aging-associated immune system dysfunction(223,224). Similarly, chronic IL-1 stimulation has been shown to promote the expansion of Cebpa-mutant HSC(225) and age-related inflammation promotes the expansion of Nras-mutant B-cell progenitors(159). Notably, recent studies have established a role for the infiltration in solid tumors of immune cells with CHIP mutations in altering the tumor microenvironment. Intratumoral CHIP is associated with reduced overall survival of patients(226,227). Therefore, in addition to directly impairing the function of adaptive immune cells which result in reduced tumor immune-surveillance, chronic inflammation has the potential to promote the expansion of malignant clones leading to increased rates of hematologic malignancies and worse prognosis of solid cancers.

Tissue normalizing pressure. Persistence and occasional progression of pre-malignant lesions with age

Natural selection has created barriers that limit the outgrowth of rogue clonal expansions with abnormal phenotypes – early lesions or non-invasive tumors. Such early lesions will often be benign but can have the potential to progress to malignancy, such as through acquisition of additional (epi)mutations that overcome such barriers, with at least one lesion doing so in about 40% of us (in the U.S. according to Cancer.gov). Fortunately, we have evolved mechanisms to prevent such progression, even if such mechanisms lose some of their efficacy at older ages. Understanding the features that dictate the relative risk that abnormal lesions will evolve towards cancer, a major goal of the PreCancer Atlas, is paramount for recommendations for screening and intervention, and the avoidance of unnecessary (and risky) procedures for lesions with minimal chance of progression(228). In this section, we will discuss premalignancies and their progression in three tissues - colon, lung and skin - given the ability of screening programs to detect early-stage disease. Similar scenarios are evident for other tissues including for the hematopoietic system, breast, and prostate, but for the sake of brevity will not be discussed further here.

Colon.

Colonic premalignancies are far more common than life-threatening cancers, indicating that most early lesions fail to progress. As revealed by colonoscopies to screen for early stage cancers, the incidence of colon polyps increases substantially with age, from 21–28% (aged 50–59 years), to 41–45% (60–69 years), to 53–58% (over 70 years)(229). Thus, polyps far outpace colorectal cancers, which have a lifetime risk of 4 to 4.5%. Interestingly, colorectal polyps are associated with biological, not just chronological age. Using DNA methylation patterns and the GrimAge algorithm, each one year of accelerated aging was associated with a 16% higher rate of detection of a preneoplastic polyp(230), suggesting that youthful tissues may limit the formation of preneoplastic polyps.

Lung.

Normal lung morphology and function decline with age, most evident after 50 years, coinciding with increases in senescent cells, chronic low grade inflammation, and the risk of lung diseases (including emphysema, fibrosis and cancer), with potential causative roles for senescent cells(231). Studies have shown progressive genomic alterations during lung adenocarcinoma evolution from atypical adenomatous hyperplasia (AAH) to adenocarcinomas(232), at the single nucleotide and chromosomal levels. Through subclonal analyses, investigators revealed that neoplastic transformation beyond the AAH stage is associated with a selective sweep of unfit clones. Of note, while cigarette smoking is clearly the dominant factor dictating the risk of lung cancer, lung cancer incidence in smokers (current and former) and never-smokers exhibit very similar age-dependence, mostly occurring in individuals over 50 (233).

Skin.

Some skin lesions, such as seborrheic keratoses (SKs), fail to progress to cancers, despite exhibiting mutations such as in PIK3CA, FGFR3, and the TERT promoter that can contribute to cancer development in other tissues and contexts(234). Actinic keratoses (AKs) are precancerous epidermal lesions, and are thought to be precursors for basal cell and squamous cell carcinomas(235). The prevalence of AKs increases dramatically in older ages, rising from 0.01% for patients between 30 and 39 to almost 15% in those over 80, with the steepest rise after 50 (236). Sun (ultraviolet radiation) exposure is the key risk factor for both AKs and skin cancers(235). Interestingly, exposure during youth appears to be important for skin cancer risk later in life(237,238), suggesting that stem/progenitor cells with initiating mutations may persist (but without progression) for many decades before evolving (in a small fraction of incidences) to malignancy. Even for cancers that can be initiated early in life, with clear causative exposures such as for ultraviolet irradiation, mechanisms are still able to limit cancer progression until older ages, a testament to the power of youth to limit the negative fitness impacts of cancers.

Metastatic suppression and dormancy

We and other animals evolved to limit the fitness impacts of malignant growths, and metastasis is the deadliest manifestation of cancers. We can consider that millions of years of vertebrate evolution have selected for mechanisms that render metastasis rare, at least until older ages (Figure 2). Primary tumors release millions of cancer cells, which must intravasate, then survive in the bloodstream, then extravasate and colonize at a new site(239). We have evolved mechanisms that make each step inefficient. Recent studies have shown how old age can promote metastatic progression (240). Changes in the ECM with age (in mouse models) have been shown to promote metastatic progression, with age-related reductions in dermal fibroblast expression of hyaluronan and proteoglycan link protein HAPLN1 enabling melanoma cell transit into lymphatics(241), and age-dependent structural changes in the collagen-rich omenta creating a permissive premetastatic niche that promotes ovarian cancer metastases in the peritoneum(242). In these studies, less metastatic disease was observed in young mice, consistent with more effective barriers in youth.

SASP factors derived from senescent cells can promote migration, invasion and metastasis of cancer cells. Senescent cells have been shown to promote carcinogen-induced skin cancer formation, growth and invasion(243) and increase the lung cancer metastatic burden from melanoma cells through secretion of soluble E-cadherin(244). Senescent mesenchymal stromal cells can alter the stiffness of the ECM and promote the migration and invasion of breast cancer cells(245) and through production of IL-6, promote breast cancer cell proliferation, migration and tumor growth in xenografts(120). Breast cancer cells themselves can become senescent and induce epithelial-mesenchymal transition, promote tumor growth and metastases of adjacent cancer cells in a paracrine manner via Notch signaling(246) or by SASP factors which are more highly expressed in breast cancer by activation of Her2 signaling(247). Similarly, a population of senescent cancer cells have been observed in the invasive front of thyroid (248) and colorectal tumors(248) where they promote migration through SASP factors leading to lymph node colonization. While senescent cells are more prevalent at older ages, further studies will be needed to determine how host age influences senescence-mediated metastatic progression.

Cancer cells can disseminate from the primary tumor early in cancer development, remaining in a dormant state for years to decades before awakening to form overt metastatic disease(249,250), with immune dysregulation facilitating dormant cancer cell persistence (e.g. altered adaptive and NK-cell responses) and awakening (e.g. enhanced inflammation). Recent investigations have shown how old age promotes melanoma cell dissemination in the skin (WNT5A-dependent), and increases metastatic progression from dormancy in the lungs (with inhibition of the WNT5A pathway)(251). Additional studies have shown how old age can promote the awakening and expansion of dormant mammary cancer cells in the murine lungs, with a critical role of high levels of PDGF-C in the aged lungs(252). From these studies, we can conversely conclude that dormancy is better maintained in young tissues. While cancer cell dormancy may seem like a mechanism for cancer cells to “lay low” for years, we can also consider that dormancy could reflect a host strategy to delay metastatic disease to years when reproductive success has historically been less likely, thus reducing the fitness costs.

Designing interventions that consider age-dependent barriers

As reviewed here, evolved barriers effectively (but not entirely) limit the burden of cancer till older ages. We argue that elderly patients would benefit from preventative and therapeutic interventions that consider these barriers and how they are altered with age. For example, understanding how RCD sensitivities change with age could guide the application of cytotoxic therapies, both for more effective elimination of malignant cells and reduced off-target toxicities. We could also consider how approaches to restore tissues and even tumor microenvironments to more youthful phenotypes could reduce cancer progression to more malignant forms.

One fundamental obstacle is that clinical trials systematically under-represent elderly individuals for a variety of reasons(253). Recently, there have been efforts to apply geriatric assessments to adjust treatment modifications and interventions in elderly patients resulting in similar effectiveness while reducing toxic side-effects(254,255). However, most proposed approaches consist in reducing the dose and intensity of chemotherapy to account for greater frailty of the patient (256) but do not consider the altered aging biology in tailored treatment strategies.

Emerging data suggest that leveraging age-specific alterations can lead to better anti-cancer strategies. One important example concerns the regulation of angiogenesis in aging. Aged dermal fibroblasts produce higher levels of secreted frizzled-related protein 2 (sFRP2), which activates the non-canonical Wnt pathway and in melanoma cells. sFRP2 induces drug resistance, angiogenesis and metastasis formation(257). Importantly, in a model of melanoma, in older mice angiogenesis is driven primarily by sFRP2 and not vascular endothelial growth factor (VEGF), consistent with the observation that the anti-VEGF antibody bevacizumab is effective in young melanoma patients but not in elderly patients(258). Another recent study looked at emerging new treatments based on nanoparticle-based drugs and found that in young mice the treatment was less effective than in old mice due to the age-related decline of macrophage clearance of the drug(259). This study is emblematic because it shows that conducting drug discovery in young settings exclusively can lead to the exclusion of potentially effective treatments. More broadly, and as discussed above, aging is associated with alterations in the immune and inflammatory microenvironment, in part due to senescent cell accumulations. Senolytic therapies have been extensively discussed as potential treatments for improving anti-cancer therapy efficacy in aged patients, but their clinical investigation is at early stages(260). Given the pleiotropic immune alterations in aging, it is surprising that immune checkpoint inhibitors show similar effectiveness in young and old patients (based on a meta-analysis in (260), recognizing variable results across trials). On the other hand, the specific altered cytokine and metabolic milieu of aged patients is rarely taken in consideration for therapeutic interventions in the elderly. While progress in this arena has been ongoing, we propose that to more effectively address cancer therapy in aging, when most cancers manifest, aging as a factor should be included in cancer biology starting from pre-clinical models all the way through clinical trials.

Summary and future directions

In humans and other animals, evolution has selected for a myriad of barriers that limit the fitness impact of cancer development and other forms of damaging somatic evolution. As we have described above, these mechanisms wane at older ages as the odds of contributing to future generations historically were reduced or became negligible. As such, aging is not linear, and nor is our risk of diseases from cancers to heart disease to neurodegeneration, which remain low into middle age with an exponential increase thereafter(261). Tissue maintenance plays a key role in limiting damaging somatic evolution, by strengthening cell competition and stabilizing selection for the evolved phenotypes, maintaining an ECM that disfavors malignant progression, and favoring robust immune surveillance at least in youth. Key molecular mechanisms of tissue maintenance and tumor suppression at the cell level include senescence and telomere attrition, RCD, (epi)genome maintenance, with other mechanisms like cell competition, controlled inflammation, and immune surveillance functioning at the tissue level. Perhaps we can use this framework to develop interventions that result in better maintenance of barriers through life, or that perhaps raise certain barriers. At older ages, some of these same pathways, such as senescence and telomere crisis, and perturbations of immune function (less adaptive but more inflammatory) can contribute to aging phenotypes and promote cancer pathogenesis. We need to recognize the tradeoffs instituted by natural selection, favoring reproductive success in youth over additional longevity or disease prevention at old ages. For example, senescence and the inflammatory pathways that are essential throughout life for host defense and tissue repair can contribute to tissue aging and carcinogenesis at older ages.

While youth is associated with self-reinforcing loops of somatic (genome/cell/tissue) maintenance, aging is associated with vicious cycles of accelerated decline (Figure 3). These cycles include i. epigenetic decay and impaired genomic maintenance(53), ii. loss of heterochromatin, ERT/ERV awakening, and interferon activation(63), iii. mutation-driven clonal expansions and inflammation(3,262), iv. cell competition with elimination of dysfunctional cells declining with the reduced fitness of neighboring cells(143,162), and v. inflammation and tissue damage(263). This list is clearly not exhaustive. These cycles will promote tissue decline with age, and likewise will increase the risks of cancer, by increasing heritable phenotypic heterogeneity, increasing inflammation, impairing anti-cancer immunity, reducing the efficacy of cell competition, and engendering tissue environments that can select for malignant phenotypes. These vicious cycles could as such contribute to our accelerating decline and exponential increases in disease risks (and death) at older ages. A key question moving forward is – how do we disrupt these vicious cycles and instead favor self-reinforcing loops of tissue maintenance?

Figure 3. Vicious cycles promoting somatic decline and tumorigenesis in old age.

Feedforward loops originating from age-associated alteration at the sub-cellular level (right) and at the cell and tissue level (left) could promote our accelerating decline and exponentially rising cancer risk at older ages. See main text for details. Created in https://BioRender.com

We know that lifestyles and exposures substantially contribute to cancer risk, and yet these cancers still exhibit incidence curves with steep increases in the later part of life (e.g. after 50 years; for example for lung(233)). Understanding how these cancers are still largely relegated to older ages, perhaps through the resiliency of tumor suppressive barriers, should be a key goal. We argue for a more holistic view of somatic decline and disease risks, including for cancer. We need to be wary of simple explanations and simple fixes for aging and cancer – these are complex processes, and there is no single causative factor nor a single metabolite or factor that can reverse them. Still, by envisioning these complex processes within the framework of life history theory – how evolution has sculpted a soma that maximizes reproductive success – we can begin to formulate a deeper understanding for how these somatic pathways of maintenance, repair and defense largely restrict cancer risk till older ages. Aging and cancer are inextricably intertwined.

Significance.

Better understanding of the multiple barriers that we have evolved to limit cancer development and how they can fail at older ages could enable the development of preventative and therapeutic interventions that boost these tumor suppressive mechanisms.