The Evolutionary Theory of Cancer: Challenges and Potential Solutions

Abstract

The clonal evolution model of cancer was developed in the 1950–1970s, and became central to cancer biology in the 21^st century, largely through studies of cancer genetics. Although it has proved its worth, its structure has been challenged by observations of phenotypic plasticity, non-genetic forms of inheritance, non-genetic determinants of clone fitness, and non-tree-like transmission of genes. There is even confusion about the definition of a clone, which we aim to resolve. The performance and value of the clonal evolution model depends on the empirical extent to which evolutionary processes are involved in cancer, and on its theoretical ability to account for those evolutionary processes. Here, we identify limits in the theoretical performance of the clonal evolution model and provide solutions to overcome those limits. While we do not claim that clonal evolution can explain everything about cancer, we show how many of the complexities that have been identified in the dynamics of cancer can be integrated into the model to improve our current understanding of cancer.

ToC blurb

Clonal evolution is now a central theoretical framework in cancer research. In this Perspective, Laplane and Maley identify challenges to that theory. There are some non-evolutionary phenomena in cancer that cannot be captured by the theory. Other challenges such as non-genetic hereditary, phenotypic plasticity, reticulate evolution and clone diversity, can be included in an expanded evolutionary theory of cancer.

Introduction

Cancers are populations of cells that are heterogeneous across space and through time. This diversity is currently a major clinical issue, limiting the efficiency of most cancer treatments as there is often a subset of cells that is resistant to whatever treatment is being used.^1,2 The diversity also limits the accuracy of prognosis and our ability to predict how it will respond to an intervention because a biopsy may not be representative of the entire neoplasm and there is a strong stochastic component to how it changes through time. A better understanding of the mechanisms involved in this diversification are thus urgently needed to better manage cancer. An explanatory mechanism of cell diversification that has gained traction in the 21^st century is clonal evolution: cancer cells diversify through the accumulation of genetic and epigenetic alterations, which can change the cellś relative fitness and consequently lead to clonal expansion or contraction by natural selection. The principles of evolution and the tools of population genetics can be and have been successfully applied to cancer cells.^3,4 However, the evolutionary view of cancer progression has been challenged⁵ and it was shown that phenotypic heterogeneity within cancers can be largely independent of the genetics of clones.⁶ Given that evolution by natural selection acts on the phenotype and relies on its heritability, these data question the relevance of the clonal evolution model. It is high time that we evaluate the strengths, weaknesses and opportunities for improving the clonal evolution model.

To achieve this, we here first explore what is known about the evolution of cancer cells through a historical and contemporary review of the literature. We then analyze the theoretical structure of the clonal evolution model by highlighting its underlying assumptions. By doing so, we intend to locate the theoretical and conceptual tensions and difficulties inherent to the current model. We conclude with suggestions and perspectives that can help address some of the issues and enhance the explanatory relevance of clonal evolution in cancer.

What is the clonal evolution model?

Historical perspective

The clonal evolution model is an abstract model depicting the evolution of cancer cells in a patient. It is often attributed to Peter Nowell, but Nowell himself always made it clear that he was summarizing ideas that were developed by the community.⁷ The observation of intratumor heterogeneity, and the idea that cancer cells evolve was already present before 1976, and its success occurred later.

In the 1950s, the notion of evolution was often used in oncology to describe cancer progression (see for example refs^8,9), though not always implying a Darwinian process.¹⁰ During this time, studies on the cancer cell karyotype supported the hypothesis first introduced by Theodor Boveri that tumors could emerge from chromosomal defects caused by abnormal mitoses, which predicted the clonal origin of cancer.¹¹ Karyotyping also led to the observation of a variation in the number of chromosomes contained by cancer cells, and, as a single cell whole genome assay, played a crucial role in revealing the evolution of cancer cells. Karyotypic observations directly raised the question of whether tumors are composed of multiple strains of cells, each having a fixed number of chromosomes, or whether the karyotype of tumor cells could change (see for example refs^12,13) and if so, whether such changes were stochastic or heritable and selected (see for example refs^14,15). T.S. Hauschka directly refers to “mutation-selection sequence analogous to phylogeny” and argues that “specific tumor karyotypes have competitive survival value.”¹⁵ Many techniques were used to resolve the question of the clonal origin of cancer, including the discovery of the Philadelphia chromosome,¹⁶ random somatic inactivation of one of the two alleles of glucose-6-phosphate dehydrogenase enzyme in women^17,18 or immunoglobulin heavy chain rearrangements.^19,20

All of those experiments converged toward the hypothesis of a clonal origin of cancer, with intratumoral heterogeneity originating from subsequent variations that may undergo selection. In most of these papers, evolutionary notions are kept in the background,¹⁰ but some directly discussed evolutionary principles applied to cancer. The English cytogeneticist C.E. Ford dedicated a paper to selective pressure in healthy, irradiated, or cancerous somatic cells defending the hypothesis that “unbalanced karyotypic changes have an effect on the probability of survival and proliferation” and thus that “the karyotypic structure of a cell population would then be the resultant of the operation of selective forces on the variability arising within it.”²¹ J. Lejeune argued that the hypothesis of selection was necessary to the coherence of the clonal evolution model and explicitly mentioned the use of karyotypes to reconstruct the “natural history of the clone”, comparing such reconstruction to “the approach of paleontologists reconstructing, from form to form, the history of the phylum”²² (p. 76, translated from French).

After Nowell’s elegant summary of the evolutionary model of cancer, we got… crickets. The evolutionary biology of cancer lay dormant for decades with only a few scientists^23–28 building on Nowell’s framework. It was not until the 21^st century that the field gained momentum. It is probably not a coincidence that interest in cancer evolution accelerated when high throughput sequencing started generating extensive amounts of genetic data from cancers. Observations of the clinical importance of intratumor heterogeneity^29–34 and clonal expansions^35–38 soon followed, along with the confirmation that therapy often selects for pre-existing clones with mutations that render them resistant to therapy.^1,39–44 Although clinical impact remains limited, further studies demonstrated that an evolutionary approach to cancer therapy can lead to dramatic improvements in time to progression and overall survival.^45–51 Amidst all this, reviews of how evolutionary biology and ecology could be productively applied to cancer^29,52–55 may have helped to lay a foundation for future progress.

How do cancer cells evolve?

Any population of entities with a diversity of heritable properties that can result in differential fitness between entities can evolve by natural selection.⁵⁶ Somatic cells are such entities and can be subject to evolution by natural selection because mutations cause heritable diversity among cells and (at least some) can alter cell fitness. The open question is whether and how much they evolve by natural selection, a question that has been debated by scientists and philosophers.^57–60

It is important to recognize that cancer cells can evolve through a plurality of mechanisms. While they can evolve by natural selection, some tumors show little evidence of natural selection and appear to be mainly evolving by neutral evolution.^61–64 A mathematical neutral evolution model is consistent with genetic data in approximately a third of solid tumors⁴, of multiple myeloma⁶⁵ and chronic myelomonocytic leukemia.⁶⁶ Another pan-analysis also highlighted that negative selection, which is predominant in the germline, is nearly absent in cancer and somatic evolution.⁶⁴ Although clonal evolution was initially considered as a continuous gradual process, it has been shown that bursts of changes can also happen.^67–70 Clonal evolution can occur through stasis, gradualism, or punctuation, with different molecular clocks ticking at different speeds,^71,72 and can proceed by linear evolution or branched evolution.^31,73–77 There is a diversity of processes that underlie the clonal evolution of cancer cells.

In addition, although the clonal evolution model traditionally focuses on the evolution of cancer (stem) cells as the unit of selection,^78,79 selection may occur at higher levels (Figure 1a), on groups of cells rather than individual cells. For example, there is a debate on whether there is evolution by natural selection between metastases, each metastasis counting as a reproductive entity, generating further secondary metastases.^59,60,80 Higher level selection could also select among colonies of cancer stem cells and their non-stem cell progeny⁸¹ or among epithelial proliferative units like colonic crypts, which may divide and die.^52,82 Selection in cancer may also occur at levels below the cell, through the evolution of transposable elements,⁸³ or extrachromosomal DNA⁸⁴ and perhaps micronuclei,⁸⁵ if they can replicate independently of cell replication (Figure 1b). The fact that selection may act at additional levels, above and below the cell, does not violate the clonal evolution model. It adds complexity to it, in the same way that multilevel selection,^86,87 above and below the level of the organism, does not violate the theory of evolution, but rather adds complexity to it.

Figure 1|. Cancer evolution by natural selection can operate at multiple levels of biological organization at once.

a) There may be natural selection on groups of cells, here shown as stem cells and their non-stem cell progeny. In this example, the blue non-stem cells have a property (marked by a ´+´) that protects their stem cells from destruction by lymphocytes, and allows the colonies of blue cells to replicate. Yellow progeny lack (-) these properties, leading to the destruction of the yellow population by the lymphocytes. b) Below the level of the cell, extrachromosomal DNA (ecDNA) may increase the copy number of cancer-related genes (for example MYC) independent of the chromosomes. Similarly, transposable elements (TE1 and TE2) in the genome may replicate within the genome.

What is the value of the clonal evolution model?

The clonal evolution model integrates current knowledge about cancer cell evolution. It is a highly theoretical framework. It aims not just at describing but also at reconstructing past evolution, predicting potential future evolution (for example resistance to treatment), explaining phenomena occurring in patients, and providing the foundation for novel therapeutic interventions. We refer to these properties as the theoretical value of the clonal evolution model.

A primary theoretical value of the model is to explain how and why cancers change over time and in response to therapy.^53,88 Phylogenetic reconstruction provides a description of the natural history of a cancer. For example, clonal mutations (also known as ´truncaĺ mutations) tend to be different from subclonal mutations (also known as mutations that only appear on a ‘branch´). Phylogenetic reconstructions can highlight changes in exposures and mutational processes over time, exemplified by the loss of the aflatoxin B1-related mutational signature in African migrants with hepatocellular carcinoma in the years after they arrived in France.⁸⁹ It also showed that contrary to expectation, metastases can occur early in carcinogenesis.^75,90–92 Lessons learned from these phylogenetic reconstructions also facilitate drug targeting (for example targeting a clonal mutation rather than a subclonal one), and help predict the risk of therapeutic resistance through selection of resistant mutations, such as the T790M mutation of epidermal growth factor receptor (EGFR) that causes resistance to first-generation EGFR inhibitors in lung cancer.⁹³ Many mutations have been associated with resistance to treatment in ways that now allow clinicians to better decide which treatment to apply. The clonal evolution model has also been used in risk stratification and prognosis, using measures of the clonal evolution, such as intratumor heterogeneity, to predict which precancers are likely to progress to invasive disease, and which cancers are likely to be lethal.^1,94–96

The clonal evolution model predicts that different environments will change the ability of mutated clones to expand. In the hematopoietic system for example, mutated clones emerge with aging, a process referred to as clonal hematopoiesis. Different environmental changes may lead to clonal hematopoiesis, such as aging, chemotherapy, infections, smoking. But each tend to select different types of mutations.⁹⁷ Clonal hematopoiesis also comes with various mutations, and various dynamics through age.^98,99 This also helps to explain the late occurrence of cancer, as the decrease of healthy cell fitness with aging provides weaker competitors for the selection of mutated clones.^100–103 Note that the promotion theory is sometime thought as in opposition to the somatic mutation theory, but these two theories are compatible in the context of clonal evolution, as fitness depends on both intrinsic cell properties and extrinsic properties.^104,105

More fundamentally, the clonal evolution model generated a profound conceptual switch that views cancer as a dynamic process. As such, it contributed to discrediting the therapeutic strategy of searching for a magic bullet and required changes in research and treatment practices. The model predicts that most advanced cancers are likely to be able to escape any treatment. This has led to various propositions for how to develop therapies based on evolutionary principles, such as drug holidays, changing drugs through time, and adaptive therapies that manage the clonal dynamics in order to prevent therapeutically resistant clones from expanding out of control.^46–49

Years of observations and experiments testing aspects of the clonal evolution model have supported its theoretical value. Yet, it does not explain everything about the phenotype of cancer. Putting cancer cells in a different microenvironment gets them to behave differently. An extreme illustration of this are the examples where placing cancer cells in a healthy environment can ńormalizé their behavior.^106–111 Phenotypic intratumoral heterogeneity may largely be independent of the genotype⁶ and phenotypic plasticity is increasingly recognized as an important contributor to intra-tumoral heterogeneity, and treatment escape.^112–115

Model assumptions and limitations

The performance of the clonal evolution model depends on two different aspects, one factual, the extent to which evolutionary processes contribute to cancer, and one theoretical, the extent to which the current clonal evolution model captures these processes. The model’s premise is that evolutionary processes do play a role in cancer. This is supported by decades of evidence. However, the importance of the model is an open question. The more evolutionary processes influence cancer, the more relevant the clonal evolution model becomes. The theoretical performance of the clonal evolution model is a different issue. The current model may only partially capture the actual involvement of evolutionary processes in cancer. It is important to separate the challenges related to the factual involvement of evolutionary processes from those related to the model’s theoretical performance. The latter can be addressed by extending the current model.

We see two important but questionable underlying assumptions in the current clonal evolution model: (1) the phenotype of cancer cells, in particular their fitness, is assumed to largely rely on their genotype (studies of clonal evolution have been historically dominated by genetic approaches); (2) the ancestral relationships between cancer cells are assumed to form a branching tree of cells, neither of which is actually necessary for a system to evolve. We will analyze both assumptions in detail, searching for limits and critiques to the model and distinguishing those that are real limitations to how much evolutionary theories can explain cancer (factual extent), and those that are invitations to improve the current clonal evolution model (theoretical limitations).

Assumption #1: fitness relies largely on genotype

The clonal evolution model draws from the modern evolutionary synthesis, which is a quantitative genetic theory of Darwinian evolution defining evolution as changes in the frequencies of genetic variants in a population through space and time. As a mathematical and theoretical framework, it requires some simplifying assumptions, in particular the organism is reduced to its genetics. In oncology, specific genetic alterations have been identified and characterized for their key functional roles in the development and evolution of the disease, which led to the success of the notion of oncogenes and tumor suppressor genes in the 1980s.^116,117 However, it is well known from developmental biology that the same genotype can produce widely different phenotypes in normal cells^116,117 and this also applies to cancer cells.¹¹⁸ The assumption that the fitness of cancer cells largely relies on their genetics faces a number of challenges.

Identifying key clonal evolution genes is challenging

Large consortiums have invested a lot of effort in sequencing tumors to identify cancer genes and several confounding factors have been identified, such as the size of the genes and their location in fragile sites.^64,119–121 The statistical significance of the presence of some mutations needs to be evaluated according to some background factors such as the mutation rate of each nucleotide, structural variations, and purity of the sample. Given that mutations in typical cancer driver genes are present in somatic healthy tissues, it also appears crucial to compare the frequency of mutations in cancer to their frequency in non-cancerous tissues. For example, NOTCH1 mutations are common in esophageal cancer and skin cancer, leading to the assumption that it is a driver gene in those cancers. However, as NOTCH1 mutations are more common in normal skin and esophageal epithelium than in the cancers of those tissues, NOTCH1 mutations may actually play a protective role against both skin and esophageal cancer.^122,123 Thus, identifying driver mutations is not trivial.

There is also a conceptual problem in identifying cancer genes.¹²⁴ The same mutation can be neutral or even deleterious in one context and become a driver in another, and vice versa. The concept of the cell of origin illustrates this phenomenon: the same mutation may have varying impact depending on which cell it occurs in.¹²⁵ Similarly, therapies change the environment of the cells which can select for clones with specific resistant mutations that expand during treatment and regress at treatment withdrawal.¹²⁶ Treatments induce drastic and rapid selective pressures, but slower changes are also possible, with the same consequences. Aging, for example, changes the fitness landscape of the cells, so that a mutation could be neutral or even deleterious in a young person but be advantageous in the context of aging.^{101–103,127} Fitness is by definition always relative to a particular environment. The environment of cancer cells changes, throughout the natural development of the disease but also due to aging, interventions and other changes in exposures. The difficulties in identifying driver mutations complicates efforts to define clones based on driver mutations (see Box 1).

Box 1 |. What is a clone?

The model of clonal evolution is based on the analysis, in time and space, of cells belonging to different clones. The concept of the clone is therefore central, but is rarely explicitly defined. There is an implicit consensus to see cancer clones as populations of cells that share a common identity inherited from a common ancestor.²⁷⁴ But what alterations people use to identify a clone differs from study to study.

Traditionally, the identity of a clone is based on the genotype, or rather parts of it. A typical definition of the clone is for example: "a set of cells that descend from a common ancestor and thus share genetic features".²⁷⁵ For practical reasons, clonal evolution studies are often based on driver mutations. Some studies use techniques allowing broader analysis of the genome such as whole exome-sequencing, or even whole genome sequencing, which does not rely on a priori knowledge of which mutations are involved in cancer. Finally, some studies favor the use of neutral mutations as an unbiased way to track clonal evolution.²⁷⁶

There is also a more fundamental ontological and epistemic challenge in the identification of a clone. The clonal evolution model relies on reconstruction of clones from incomplete information: we neither have access to all the cancer cells, nor to all the heritable properties of each single cell. This can lead to ćlonal illusioń, when a subclonal mutation appears to be clonal because a biopsy was only taken within the subclone.²⁷⁷ It also relies on choices made by investigators as to which alterations are relevant for defining a clone and which can be ignored.¹²⁴ Changing the list of mutations that are deemed relevant and thus used to reconstruct clonal evolution affects that reconstruction. The number of clones identified, for example, will depend on the number of genes studied as well as the depth of sequencing.⁹⁵ As an extreme illustration of this issue, one may define a clone as the set of cells that are genomically identical. However, whole genome sequencing at the single-cell level would likely reveal that each individual cancer cell is unique, and therefore count as a new clone, making the concept of a clone useless.²⁵⁹ The history of the clonal evolution model shows that these choices of what alterations should be used to define a clone have changed through time. This historical contingency highlights the arbitrariness in definitions of clones. In a pluralistic approach, one could argue that which clone delineation is the right or best one will depend on the perspectives and aims one adopts. Whether, and how much, clone delineation is an issue largely depends on what is expected from the clonal evolution model. It is for example an issue for any study that quantifies intratumoral heterogeneity by counting clones. It is much less of an issue for those who may want to reconstruct phylogenetic relationships between various cancer cells (for example between metastases at different locations).

There are some alternative views of clonal identity: several studies have now been using epigenetics in their reconstruction of cancer evolution.^{27,37,129,143,144,147,278} A few have supported a phenotypic or functional clonal identity. For example, a clone was defined as "a group of cells with the same phenotype, which have expressed that phenotype consistently since their most recent common ancestor".⁶³ That definition might call a set of stem cells a clone if they shared the same phenotype and had a recent common ancestor, but would exclude all the non-stem cells that were part of the same branch on the cell lineage tree. In contrast, in developmental biology, a clone has been defined as “the in vivo descendants of a single ancestral cell", equivalent to a monophyletic clade in a phylogeny.²⁷⁹

Under every cancer is a cell lineage tree describing the relationships and history of the cells within the cancer.^74,168 Defining clones is a matter of dividing up that evolutionary tree (Figure 2).

Genes are not the only inherited material

It is now well appreciated that many epigenetic properties are heritable at cell division and can contribute to clonal evolution,^128–137 including the evolution of metastasis^138,139 and resistance to drugs.^133,140,141 However, epigenetics encompasses a broad range of phenomena, some of which are random and heritable, while other are under the control of the cell and may change in response to some form of signaling, for example during differentiation. The heritable alterations could be integrated in clonal evolution as epimutations.¹⁴² Studies have found a concordance between genetic and epigenetic clonal evolution, which makes sense given that there is one unifying lineage of the cancer cells.^143–145Additionally, there can be multiple epigenetic subclones within one genetic clone, or vice versa.^146,147 This again raises the question what information to use to delineate and count clones (Box 1, Figure 2). Epigenetic alterations challenge the theoretical accuracy of the clonal evolution model when the model relies only on genetic mutations, but it does not challenge its factual extent.

Figure 2|. Defining clones is a matter of dividing up the cancer cells tree.

a, The tree of cancer cells, with genetic, epigenetic, and phenotypic information overlaid. b, The cancer cells tree with clones identified through driver mutations. 3 clones are identified (white with whatever mutations the transformed cell had, blue with the occurrence of the blue mutation b, turquoise with the turquoise mutation t). c, The cancer cells tree with clones identified through epigenetic alterations. 3 clones are identified (white with the epigenetic state of the first transformed cell, red with epigenetic mutation r, yellow with epigenetic mutation y). They are not the same clones as in the genetic clones. d, Phenotypic clones. There are 2 cell types. The turquoise genetic clone is mixed and contains both cell types, while the white and blue genetic clones contain a different cell type. The yellow epigenetic clone is mixed, while the red and white epigenetic clones contain the same cell type. Are there 3 clones, 5 clones, or more?

Heritable properties may extend beyond the genetic material and its epigenetic alteration. As cell reproduction happens through cell division, daughter cells inherit a portion of the cytoplasm of the mother cell, including organelles, RNA, proteins. They also inherit the microenvironment that the mother cell may have modified.^213–215 This may generate short-term heritability of the cells’ phenotype with time scales depending on the fluctuations and turnover rate of those molecules and the microenvironment. In bacteria, the use of śister machineś that separate the two daughter cells in independent growth channels, revealed that different non-genetic traits present different ´memorý patterns under different time-scales, from 2 to 10 generations.¹⁴⁸ In a melanoma cell line, 227 genes that show transient heritable high expression (around 40h, sometimes reaching 5 days) were identified,¹⁴⁹ of which 162 were previously associated with resistance to therapy.¹⁵⁰ Cells sorted for high expression of two of these transiently heritable genes, specifically EGFR and nerve growth factor receptor (NGFR) showed a much higher level of resistance to the MEK inhibitor trametinib than unsorted cells.

The inheritance of properties at different time-scales remains largely ignored, however given the short duration of therapeutic interventions, these alternative forms of inheritance may be of clinical relevance. As discussed below, improving our understanding of inheritance might improve the theoretical performance of clonal evolution.

One genotype can take on many phenotypes

Normal somatic cells in multicellular organisms are phenotypically plastic. Phenotypic plasticity has long been studied in the field of development and is increasingly recognized as important in cancer as well. From a single genotype, cells can take on different phenotypes as a result of at least three non-exclusive processes: differentiation, response to other extrinsic signals, and stochastic fluctuations (of gene products and epigenetic modifications). The latter two can include dedifferentiation, transdifferentiation, or other changes in cell states such as metabolic changes. Cancer cells inherit and modify the phenotypic plasticity of the normal cells from which they evolve.

Currently, phenotypic variations that are independent of the genotype are disregarded by the clonal evolution model. Thus, plasticity is taken as a challenge to the factual extent of the clonal evolution model: the more non-heritable plasticity is causally involved in cancer, the less evolutionary processes are. For example, a study of dormancy and relapse in estrogen receptor–positive breast cancer treated with adjuvant endocrine therapies showed that the treatment stochastically induces a dormant state, which was unstable, in a fraction of cancer cells from a random diversity of clones.¹³³ In such cases, evolution by natural selection does not play much role in cancer cells ability to survive treatment. However, conceptual fuzziness about phenotypic plasticity obscures the debate.¹⁵¹ Some cases of plasticity might heritable and in fact only be a theoretical challenge which may be incorporated to improve the clonal evolution model.

First, the phenotype of cancer cells often relies on non-genetic, yet heritable, properties, such as epimutations. Such epigenetic-driven phenotypic diversity can easily be integrated in the clonal evolution model by incorporating epimutations. The same should apply to any other non-genetic heritable property that contributes to the cell phenotype.

Second, the degree of phenotypic plasticity can be heritable, encoded in the genome or epigenome (reviewed in ref¹⁵¹). Certain cell lineages may be more plastic than others, providing them an advantage in changing microenvironments. Other cells have phenotypes that are remarkably stable, such as fully differentiated cells. Some (epi)mutations may increase cell plasticity, and selection on those (epi)mutations will flow indirectly through direct selection on the phenotypes they produce. There is relevant literature on the evolution of mutation rates^152–155 and the evolution of phenotypic plasticity^156–158 that might be used to integrate phenotypic plasticity into the clonal evolution model.

Third, phenotypic plasticity, includes a variety of phenomena such as differentiation (a robust, channeled, predictable process in multicellular organisms), response to extrinsic signals (also partly predictable, although conditional on microenvironmental signals), and stochastic fluctuations (unpredictable). The differences between these processes matters if one wants to address the empirical limitations of evolutionary accounts of cancers. Why these types of plasticity need to be conceptually distinguished, can be illustrated by two hypothetical contrasting examples. First, when a treatment selects stem cells due to their upregulation of efflux pumps, selection for efflux pumps will maintain the clonal composition. This is because all genetic clones contain cancer stem cells, although there may be changes in clone size based on the proportions of cancer stem cells in the different clones.⁸¹ Second, when a cell cycle-specific treatment selects for a slow-cycling or non-proliferating cells, and if quiescence is induced by a particular microenvironment, then clones that happen to be in that microenvironment will survive. These two cases present different challenges to the clonal evolution model. In the first case, the maintenance of all genetic clones may hide the selection of stem cells. In the second case, resistance might wrongly be attributed to the genetics of the clone when it is its location that caused the resistance. Other scenarios involving plasticity may also challenge the clonal evolution model. For example, the clonal evolution model takes the cancer stem cells to be the units of selection,⁷⁹ but dedifferentiation may occur^159,160 allowing non-stem cells to contribute to the evolutionary dynamics.^81,161

While phenotypic plasticity in response to external stimuli represents a major limitation to the factual involvement of clonal evolution, differentiation-based and stochastic-based types of plasticity represent theoretical challenges. As far as differentiation remains robust and predictable, this type of plasticity could be included in mathematical models of clonal evolution.⁸¹ Similarly, stochastic plasticity is amenable to mathematical modeling that can be introduced in evolutionary models of cancer.¹⁶²

Many genotypes can produce the same phenotype

Such phenomena can be cases of convergent evolution, phenotypic plasticity, or non-heritable factors. Selective pressures may select for the same phenotype in independent lineages, which is well accounted for by the evolutionary models of cancer. We have argued that the hallmarks of cancer are cases of convergent evolution that occur due to natural selection acting on unrelated cells.⁵⁸ This challenge to the genetic evolutionary model of cancer led to the hypothesis that we gain more clarity and control by focusing on the evolution of phenotypes.¹⁶³ As we have argued above, differentiation processes and responses to environmental signals can result in similar phenotypes regardless of the underlying genetics. There are other non-heritable determinants of phenotype that may be independent of the clones genetics. For example, cancer cells can transfer proteins, lipids and nucleic acids via extracellular vesicles which can contribute to the malignant phenotype.¹⁶⁴ In a mouse model of gliomas, it was shown that cancer cells with mutated EGFR can transfer the truncated oncogenic form EGFRvIII to cancer cells lacking the mutation, leading to the acquisition of the mutated phenotype by these cells.¹⁶⁵ Distinguishing between heritable and non-heritable determinants of phenotypes is key, as the second is a direct challenge to the factual extent of the clonal evolution model, although evolutionary theory can account for non-heritable determinants of phenotype as a form of noise, by including measures of the heritability or evolvability of traits in models of phenotypic evolution.^166,167

Assumption #2: A cancer cell lineage is a tree

The clonal evolution model assumes a tree shape of the cell lineages,¹⁶⁸ where cell lineages divide and die but do not fuse. Clones are defined by descent from a single common ancestor cell (Box 1). Inheritance is vertical, from the parental cell. Moreover, basing the clonal evolution model on cell lineages implicitly assumes that cells are the only unit of selection. Several observations question these assumptions.

Cell fusion and cancer cell modes of reproduction

Cancer cells can fuse into viable, reproducing cells that can give rise to clonal populations of cells.^169,170 Cell fusion in cancer is considered rare and has been largely neglected. Yet, several studies reported fusions between cancer cells and non-cancer cells (hybridomas being a famous case of such fusion), or between different cancer cells. In patients, fusions have been reported with hematopoietic cells in the context of bone marrow transplantations (BMT) that offer markers to track cell fusions.^169,171,172 As an example, a mother who underwent BMT from her son, later developed a renal carcinoma with a fraction of the cancer cells containing both chromosomal alterations of the cancer cells, and a Y chromosome from her son’s cells.¹⁷¹ Another study analyzed tumor biopsies from 7 patients with various cancers (pancreatic ductal adenocarcinoma, renal cell carcinoma, head and neck squamous cell carcinoma, and lung adenocarcinoma), and found evidence of fusions between cancer cells and leukocytes in all of them.¹⁶⁹ Following up on this result, they searched for circulating hybrid cells in patients with pancreatic cancer and found that the quantity of circulating hybrid cells correlated with advanced disease and was associated with poor prognosis, which was not the case of non-fused circulating tumor cells, raising the possibility that fusions may increase the risk of cancer progression.

The role of cell fusion in cancer cell evolution has been studied in vitro and in animal models. Hybrids of cancer cell lines and macrophages, once transplanted into mice, had a shorter doubling time than the maternal cancer cell line, suggesting that cell fusion can increase the fitness of the cells.¹⁶⁹ Cell fusion can also provide cancer cells with new properties. For example, fusion with hematopoietic cells such as macrophages can endow the fused cancer cells with migrating abilities.^{169,173–177} Several studies discuss fusion as a possible alternative mechanism to (but not mutually exclusive with) the Darwinian clonal evolution explanation of metastasis according to which the ability to metastasize is gained through a process of evolution by natural selection.^176,177 Other studies have shown that fusion between cancer cell lines can generate new cell lines that are more malignant and therapeutically resistant than the original cell lines.^169,170 Cell fusion has also been discussed as a possible mechanism to produce cancer stem cells.^178,179 Last but not least, cell fusion is a mechanism that can generate diversity through subsequent stochastic loss of genetic material.¹⁸⁰

Cell fusions do not undermine the factual extent of clonal evolution but add a level of complexity regarding processes involved in their evolution that is not currently taken into account by the clonal evolution model.

Horizontal gene transfer

The clonal evolution model assumes that inheritance is vertical with cell division. But several mechanisms of horizontal gene transfer have been described such as exchange of mitochondria and DNA. Co-culture of a lung cancer cell line depleted of mtDNA with bone marrow non-hematopoietic cells or skin fibroblast,¹⁸¹ or transplantation of breast cancer cell lines depleted of mtDNA¹⁸² showed rescue of mitochondrial function through mitochondrial transfer. Introduction of mitochondria from mesenchymal stromal cells (MSCs) into cancer cells led to an increase in oxidative phosphorylation, ATP production, migration and proliferation of the cancer cells.¹⁸³ Horizontal transfer of mitochondria through tunneling nanotubes from endothelial cells to MCF7 breast cancer cells was shown to improve chemoresistance to doxorubicin in vitro.¹⁸⁴ Similar results were also obtained through co-culture with bone marrow stromal cells in vitro or after engraftment of acute myeloid leukemia blast cells from patients or cell lines in mice.¹⁸⁵ Coculture of glioblastoma stem cells (GSCs) and MSCs also led to metabolic rewiring of GSCs following mitochondria transfer from MSCs to GSCs, increased proliferation and resistance to temozolomide in GSCs.¹⁸⁶

Horizontal gene transfer can also occur in cancer due to the uptake of DNA from extracellular vesicles, including apoptotic bodies from dying cancer cells.¹⁸⁷ This has been shown in vitro^{187,188 180,181} as well as in mouse experiments,¹⁸⁹ and can be prevented by treatment with DNAses.¹⁸⁹ Microvesicles have been shown to carry oncogenes such as MYC or HRAS^V12 ref^188,190,191 that can be taken up by other cells. The extent and importance of horizontal gene transfer in cancer through extracellular vesicles remains an open question.

Horizontal gene transfer does not contradict an evolutionary model of cancer. Rather it adds complexity to the model. Adaptation through natural selection can occur through horizontal transfer in addition to vertical inheritance.

Trogocytosis and horizontal transfer

Beside gene transfer, phenotypic properties can also be horizontally transferred through trogocytosis, a mechanism of membrane fragment transfer between cells. Trogocytosis is a relevant phenomenon for cancer cell evolution as it allows cells to acquire new phenotypic properties independently of their genetics and epigenetics, leading to a mismatch between the genotype and phenotype. For example, in a colon cancer mouse model using patient-derived xenografts, cancer cells were able to acquire lymphocyte membrane proteins including lymphocyte cellular markers, as well as immune regulatory surface proteins, which suppress activation of immune cells.¹⁹² Similar results were obtained using a mouse model of leukemia,¹⁹³ although in this study, Natural Killer (NK) cells and CD8⁺ T cells acquire the checkpoint receptor programmed cell death protein 1 (PD1) from leukemia cells, which results in the suppression of NK cell antitumor immunity. Trogocytosis has been implicated in chimeric antigen receptor-T cell (CAR-T) cell and CAR-NK cell escape.^194,195 Trogocytosis can only be meaningful to clonal evolution if the phenotype it induces can be transmitted through at least one cell division. Cell surface materials are partly maintained through cell division, but the symmetry of inheritance to daughter cells is unknown, as well as the number of divisions after which the phenotype is lost.

Cell cannibalism is another process that can provide properties to cells that change their fitness under particular selective pressure. For example, metastatic melanoma cells can cannibalize T lymphocytes, which allow them to survive under serum deprivation.¹⁹⁶ Similarly, a breast cancer cell line was shown to cannibalize mesenchymal stromal cells, which also enhanced their survival in the context of starvation, through the induction of a dormant state.¹⁹⁷ Again, its contribution to clonal evolution depends on the heritability of those phenotypes.

Horizontal transmission of phenotypes due to trogocytosis or cell cannibalism reduces the heritability of those phenotypes and so reduces the factual extent of the clonal evolution model.

Intracellular microbes

Although clonal evolution refers to the evolution of cancer cells, presence of microbes (such as viruses, bacteria and fungi) inside cancer cells may call for a multispecies view of clonal evolution.¹⁹⁸ Viruses have been implicated in carcinogenesis for a long time, and in cases where they directly cause carcinogenesis and even integrate into the cancer genomes, they become part of the genetic material that is evolving in cancers.^199,200 Recent studies have observed that bacteria and fungi are not just around some cancer cells, but that they can also reside inside them^201–205 and it was even suggested that most intratumoral bacteria are intracellular, residing in both cancer cells and immune cells.²⁰²

Intracellular microbes can change their host cell functional properties and fitness, thereby contributing to clonal evolution. They can increase the fitness of cancer cells, for example, in colorectal cancers, Fusobacterium nucleatum have been associated with increased tumorigenicity.^206,207 Fusobacterium-positive tumors engrafted in patient-derived xenografts models while Fusobacterium-negative ones did not, a difference that was reversed by antibiotic treatment.²⁰¹ Bacteria can inhibit or elicit immune responses, for example through peptide presentation^203,208 and can also mediate cancer cells resistance to therapies, for example through metabolization of chemotherapies into an inactive form.²⁰⁹ Human Papilloma Virus (HPV) interferes with multiple tumor suppressor mechanisms and increases the fitness of the neoplastic cells that they infect.²¹⁰

Thus, intracellular microbes may contribute to clonal evolution in predictable ways that could be implemented in the clonal evolution model. The heterogeneous and focal distribution of intracellular bacteria lead to the notion of ´microbial intraclonal diversitý whereby different fitness of cancer cells of the same genetic clone is dependent on the presence of intracellular bacteria.¹⁹⁸ There are, however, theoretical challenges to the integration of the role of microbes in clonal evolution. One of them is the mode of transmission of these microbes. They may be transmitted vertically, and thus simply represent a subclone of infected cancer cells, as well as horizontally by transiting from one cell to another. In the second case, they question the tree-shape of inheritance patterns among cancer cells. Moreover, bacteria themselves may undergo their own evolutionary dynamics. Bioinformatic analysis of the metabolic activity of intratumoral bacteria suggested that bacteria are under their own selective pressures shaped by the tumor cells and their microenvironment.²⁰² Thus various selective processes may act on cancer cell lineages and microbial lineages.¹⁹⁸

The evolution of cancer cells may thus not be a simple branching tree of cells, but might involve fusion and transmission between branches (Figure 3). Genetic and non-genetic inheritance may not just be vertical but may also sometimes be horizontal. Additionally, clonal evolution might not only concern cancer cells but also intracellular bacteria and fungi that may change the fitness of cancer cells. These mechanisms challenge the theoretical efficacy of the current clonal evolution model rather than the factual extent of the involvement of cancer cell evolution. They thus open avenues to improve the theoretical power of the clonal evolution model.

Figure 3|. Clonal evolution is more complex than a bifurcating tree of cells.

Clones can use horizontal gene transfer (dotted arrows) to transfer heritable information, thereby breaking the assumption of vertical inheritance. The purple bean-shaped cells represent intracellular microorganisms, which evolve and have their own lineages indicated by the purple dotted lines. Elements of the cells can also be transmitted horizontally from cells, including non-malignant cells, to cancer cells, through absorption of extracellular vesicles, tunneling nanotubes, trogocytosis or cell cannibalism, which is represented by the transfer of cell surface receptors from the blue cell (for example a lymphocyte) to the gray cancer cell.

Improving the theory of cancer

The above challenges and limitations of the clonal evolution model suggest opportunities for improving the model. We will focus on four potential extensions to the clonal evolution model: the diversity and time scale of inheritance beyond genetics, phenotypic plasticity, reticulating modes of evolution, and the concept of clone.

Inheritance and time scale

Inheritance is a central concept for any evolutionary theory of cancer. There cannot be evolution without some sort of inheritance. But inheritance is loosely defined, and its definition is subject to debate.

Several classifications of forms of inheritance have been proposed.²¹¹ This includes the distinction of internal from external channels of inheritance.²¹² Internal inheritance comprises factors that are transmitted through cell division (or cell fusion) such as genetic and epigenetic inheritance. External inheritance contains every transmission that is not passed through cell division. Internal inheritance can be traced through cell lineages while external inheritance cannot.

It is relatively easy to extend genetic models of inheritance to epigenetic models, at least for randomly mutating epigenetic states.^143,213,214 But internal inheritance can also extend beyond genetics and epigenetics. One approach to studying non-genetic and non-epigenetic heritable phenotypes could be to quantify the heritability of those phenotypes using the breeders’ equation.²¹⁵ By measuring change in cancer cell phenotype (in vitro or in vivo) over different timescales and selective pressures, cancer biologists could quantify the degree of heritability of a phenotype of interest. This would account for both internal and external modes of inheritance, including non-(epi)genetic inheritance of the cytoplasm and niche construction. Note that these measures of heritability depend on the degree of heritable variation in the population. A phenotype could be heritable but would appear to be not if the population did not include variation in the substrates that encode that phenotype.

There is external inheritance in cancer, at least through niche construction.²¹⁶ There is a form of co-evolution between organisms and the niches that they alter, that can be represented in formal models of evolution.²¹⁷ When neoplastic cells engage in niche construction, by activating fibroblasts,^218,219 recruiting immune cells,^220–222 inducing angiogenesis,²²³ and otherwise altering their microenvironment,^224–227 they generate a form of external inheritance that changes the selection pressures on themselves, and thereby change their own evolutionary trajectories.²¹⁷ The timescale of such external inheritance can be long. For example, activated fibroblasts can maintain their pro-tumor phenotype even in the absence of cancer cells, through autocrine loops, epigenetic alterations, and even genetic alterations.²²⁸ Efforts have been made to start integrating ecology in the clonal evolution model,⁹⁴ but accounting for niche construction remains challenging.

There have also been debates about which kind of inheritance matters more.^229–231 One argument is that there is a causal asymmetry between factors transmitted over many generations and those transmitted over one generation, the former playing a more substantial role in evolution because they can accumulate over long periods of time by natural selection.²³² In cancer, however, selective pressures change at very different time scales. Aging slowly reshapes the selective landscape over decades.^101–103 The hallmarks of cancer, that are necessary for transformation, must last for the time span of neoplastic progression, which can take decades.¹⁴⁷ Short-term forms of inheritance cannot meaningfully contribute to these long processes of evolution by natural selection.²³² However, there are other barriers to neoplastic progression, such as survival in circulation during metastasis, which likely only needs to be overcome for a matter of hours or days. In that case, inheriting some key proteins from a parental cell, or a niche, like a cluster of neoplastic cells that may safely travel through the bloodstream, may last long enough to deliver the neoplastic cell to a new metastatic microenvironment. Similarly, treatments like radiotherapy and chemotherapies, which impose intense selective pressures on cancer cells, are often applied for the duration of only a few cell generations, similar to the time-scale of cytoplasmic inheritance.

The focus on genetic clonal evolution can only partially capture the short time scale events of evolution and several avenues could help improve the model in this regard. One approach involves reconstructing clonal evolution using a variety of molecular clocks. DNA, copy number variants (CNVs), and single nucleotide variants (SNVs) mostly change over years.^233,234 Methylation of CpG sites have faster clock rates than SNVs, although those rates may vary from site to site in the genome.¹⁶² Other epigenetic modifications to histones^235,236 or changes to cell state through signal transduction and transitions to new attractor states in the genetic regulatory network^237,238 can occur rapidly and may require their own molecular clocks in clonal evolution models.

Future research should characterize the time scales of different mechanisms of inheritance (Figure 4). Any phenotypic change that lasts less than the cell lifespan cannot be a meaningful heritable property for clonal evolution. Any phenotypic change that lasts less than the time scale of a selective pressure will have little effect on the cell population’s evolution in response to that pressure. In contrast, mechanisms of inheritance that produce phenotypic changes that last as long or longer than selective pressures, will be important to a population’s response to these pressures, as long as there is relevant variation in the cell population. There are mathematical methods for disentangling the different forms of inheritance, including genetic, epigenetic and ecological inheritance.²³⁹

Figure 4|. Types of inheritance over various timescales.

Genetic inheritance may contribute at any time scale and persist throughout life. Niche construction by cancer cells may have selective effects lasting for days to months, though the exact timescale is unknown. Some entities such as cytoplasmic and cell surface proteins may be inherited for a very short period of time, likely just days. They may be important contributors to evolution by natural selection in case of abrupt and short changes in selective pressures such as cancer therapies.

Phenotypic plasticity

Phenotypic plasticity can weaken the relationship between genotypes and phenotypes, reducing the heritability of the phenotype, and thereby dampening the effects of natural selection. However, phenotypic plasticity itself can be an adaptive phenotype and has evolved multiple times by natural selection.^156–158 At least one historical theoretical framework supports a contribution of phenotypic plasticity to genetic evolution: the Baldwin effect, a process by which an initial phenotypic accommodation is later reinforced by genetic adaptation.

The Baldwin effect, in its simplified form,^240,241 refers to a three-step process: first, some individuals (in our case, cancer cells) accommodate environmental changes through phenotypic plasticity. This initial phase of phenotypic plasticity allows some individuals to avoid extinction. Second, some genetic (and thus heritable) changes occur in the surviving population that allow them to survive in the new environment without having to pay the cost of phenotypic plasticity. Third, these genetically determined phenotypes are favored by natural selection and finally spread in the population because they achieve the same or more fit phenotypes as the phenotypically plastic individuals without having to pay the cost of plasticity, which may be a metabolically demanding process. Hence, the Baldwin effect is a particular case where an adaptation is initially individual and non-heritable and becomes hereditary and selected.

Several studies suggest mechanisms of initial phenotypic adaptation that are secondarily genetically selected.^133,242,243 For example, when exposed to EGFR inhibitors, most EGFR mutated lung cancer cells die, but a few can survive through a phenotypic accommodation called ´persisteŕ state. Sequencing and drug screening of relapsed clones developing from persistent cells of a single initial EGFR mutated clone revealed secondary selection of heterogeneous resistance mechanisms,²⁴² suggesting a case of Baldwin effect. Similarly, it was observed that the EGFR^T790M resistance mutation is not present at the time of therapy in some cases and can evolve from persister cells.²⁴³ Persister cells need not be quiescent. Brief exposure to vincristine can activate efflux pumps which protect cells from the drug, allowing them to persist in an active state. That activation is stably inherited by epigenetic modifications⁵ which may later be stabilized by genetic alterations.

The Baldwin effect has long been neglected in evolutionary biology as a minor phenomenon that simply ´buys timé,²⁴⁴ and thus is of interest only in cases where selective pressures change too fast for the traditional variation and selection process to produce adaptation²²⁶ Cancer is precisely one of those situations where selective pressure may change quickly (for example the onset of treatment) and buying time might be a crucial evolutionary mechanism that could be incorporated in the clonal evolution model.

Reticulation, introgression and open lineages

Cell lineages are usually assumed to form a tree, but as discussed above, cancer cell fusion can occur, and inheritance can also be horizontal, with possible transfer of mitochondria, extrachromosomal DNA, pieces of cells such as membranes, or intracellular bacteria and fungi. Those processes all violate the tree assumption and impact phylogenetic reconstruction, as the cell tree not necessarily matches the gene trees (Figure 3). It also impacts clonal deconvolution of bulk assays,^245,246 as horizontal gene transfer and cell fusion violate many of the assumptions those algorithms are based upon. Multiple discordant gene trees may exist within a cell lineage, reflecting the diverse flows of genes through all these independent phenomena. Similar lineage violations are common in prokaryotes, giving rise to the concept of reticulated evolution. Reticulation has led many to argue that the tree of life is a misrepresentation.^233,234,239 It may be more appropriate to represent evolution as a network or web of interlaced branches.^247–249 Some suggest replacing ćlades ´ by ćlans ´ when there is no clear unique common ancestor, to avoid unjustified assumptions coming with the traditional terminology of phylogenetics.²⁵⁰ This could apply to clones. Several measures have been developed and used such as the concordance factors,²⁵¹ diversity indices for unrooted trees;²⁵² and new computational methods and statistics have been developed to identify lateral transfer and analyze phylogenetic networks.^253,254 In general, the revised view of evolution that developed around reticulated evolution calls for a multi-level perspective, sometimes also referred to as integrative evolution,^252,255 where multiple genetic worlds contribute to the observed diversity.²⁵⁶

These conceptual and computational tools represent opportunities to improve our understanding of clonal evolution. Given that horizontal genetic transfers can provide evolutionary short-cuts to acquire functional adaptations such as the ability to metastasize or multi-drug resistance, it is important to evaluate the implications of such reticulation events in cancer progression, and to adopt an integrative clonal evolution model.

Quantifying Clonal Diversity

Counting clones as a proxy of intratumor heterogeneity and cancer cell evolution is based on two implicit assumptions: that the cells within a clone are homogeneous (intra-clonal homogeneity) and distinct from other clones (inter-clonal heterogeneity), so that their number reflects the diversity of the tumor cell population. But as discussed above, these assumptions might be disputable (Figure 2). If we use mutations to delineate clones (Figure 2a), there are 3 clones and thus 3 relevant populations of cancer cells with distinctive characteristics. However, there is a different delineation of clones with epigenetics (Figure 2b), and their integration (Figure 2d) suggests that there are more than 3 relevant populations of cancer cells.

One can measure the degree of intra-clonal homogeneity and inter-clonal heterogeneity. This could be done by measuring the distance between any 2 cells inside and between clones for any phenotypic or genotypic measure. For example, a measure of transcriptomic diversity inside barcoded lineages in three mouse models of acute myeloid leukemia²⁵⁷ observed that clones are clearly not homogeneous, as single cell RNA sequencing often reveals smears of cells across phenotype space, rather than tight clusters. In contrast, another study observed high intra-clonal homogeneity inside barcoded lineages and inter-clonal heterogeneity between barcoded lineages in a human melanoma cell line.²⁵⁸ Single cell genomics analysis can also be used to measure clonal diversity.^{168,259–264} One analysis found evidence for the existence of just a few clones, with little variation within clones and a large degree of differences between clones based on copy number profiling of two breast cancer cases.¹⁶⁸ In contrast, the same team found massive genomic diversity in point mutations in two other breast cancer cases, leading the authors to question the concept of a clone as no two tumor cells were found to be genetically identical.²⁵⁹

It is possible to abandon the counting of clones altogether. As they are a proxy for measuring the structure and degree of genetic diversity in a neoplasm, we may replace the counting of clones with other methods of characterizing that diversity. For example, if the evolutionary relationships between samples can be reconstructed into a tree, then there are a variety of tree statistics that may actually be more sensitive and representative of the genetic diversity than the number of clones.²⁶⁵ In fact, some of these measures provide information on the population dynamics and selection within the neoplasm.²⁶⁵

Conclusions

There is no doubt that somatic cells undergo evolutionary processes, including both genetic drift and natural selection. They acquire heritable alterations, some of which affect their reproduction and survival. The fact that cancer cells evolve was realized in the 1950s and mostly investigated since the 21st century. However, how much evolutionary processes explain cancer remains open to debate. The success of the evolutionary theory of cancer for explaining many of the phenomena of cancer, and for developing better interventions, clearly shows the utility of the theory. However, it is also clear that non-evolutionary mechanisms contribute to cancer. The field is currently gauging the respective role of evolutionary and non-evolutionary mechanisms in cancer. But this cannot be done if we lack clarity on the boundaries between evolutionary and non-evolutionary mechanisms. There are many theories and models of cancer today²⁶⁶ whose integration is highly challenging.²⁶⁷ It’s unclear which ones are compatible or not with evolutionary theories. In general, other theories of cancer, like the cancer stem cell theory,²⁶⁸ the atavistic theory,^269,270 and the wound that will not heal,^271–273 describe the biological constraints and affordances within which somatic cells evolve. To address this issue, we focused on the boundaries of evolutionary processes, distinguishing two kinds of boundaries: theoretical and factual (Figure 5). The factual boundaries of the evolutionary dynamics of cancer extend beyond the current theory and the evolutionary theory of cancer needs to be expanded to account for phenotypic plasticity, alternative modes of inheritance, horizontal gene transfer, cell lineage fusions, and better measures of clonal diversity. Much more remains to be discovered about those phenomena and their relative importance in cancer. An improved evolutionary theory of cancer should lead to improvements in our prediction, prognosis, and treatment of cancer.

Figure 5|. Explanatory power of clonal evolution.

When aiming to explain cancer in its entirety, evolutionary processes can explain only some, but not all cancer phenomena. Other theories, such as the cancer stem cell theory or some versions of the persister cell models provide complementary explanations for various phenomena in cancer. The clonal evolution model only captures parts of the causal role of evolutionary processes in cancer. There are thus opportunities to increase our current ability to provide evolutionary accounts of cancer, by integrating heredity, plasticity, reticulate evolution and clone diversity into the clonal evolution model. The proportion of cancer phenomena that can be explained by evolution is unknown.

Glossary terms

chronic myelomonocytic leukemia

A type of chronic leukemia affecting the myeloid cells and defined by a persisting monocytosis.

stasis

long periods during which no or few evolutionary changes occur.

convergent evolution

independent evolution of a similar phenotype in different lineages.

breeder’s equation

ΔZ=h²S describes the change in the mean phenotype (Z) in the population as a function of its heritability (h) and the selection pressure (S) placed on that phenotype.

clades

A monophyletic group, i.e. a group of all the individuals that derive from a common ancestor.

Clans

an unrooted analogue of monophyletic group or clade, coming from the Gaelic for family. This definition comes from reference²⁵⁰.

epistemic challenge

Challenges that relate to our knowledge.

intratumor heterogeneity

Variation between cells within a neoplasm.

Hybridomas

Cell lines coming from the fusion of B cells with immortal myeloma cell lines that are used to produced monoclonal antibodies

karyotype

The set of chromosomes of a cell, including any large scale abnormalities visible in a mitotic spread.

clonal mutation

A mutation that derives from a single ancestral cell and is shared among all its descendants.

subclonal mutation

A mutation that is present in only a subset of the cells within a clone.

clonal expansion

An increase in the number of cells deriving from a common ancestor that defines the clone.

fitness

The ability of a cell to survive and proliferate in its current microenvironment.

neutral evolution

Changes in allele frequency in a population due to chance rather than fitness.

negative selection

The removal of mutations from a population due to their negative effect on the fitness of the organism or cell.

gradualism

Evolution through slow continuous small changes in the phenotypes of organisms over time

punctuation

Evolution through sudden important changes in the phenotypes of organisms

molecular clocks

Molecular processes that change proportional to time and can so be used to infer time since past events.

linear evolution

Evolution characterized by a sequence of fixation events where one genotype is replaced by another in the population, such that change over time can be described by a single sequence of genotypes.

branched evolution

Evolution in which lineages split such that multiple lineages coexist, and can be represented as a tree.

epithelial proliferative unit

An organizational sub-structure of epithelial tissues consisting of one or a few stem cells along with their partially and fully differentiated progeny.

phylogenetic reconstruction

The process of inferring the ancestral relationships between organisms or cells.

monophyletic clade

The set of all the species that descended from a common ancestor.

driver gene

A gene in which particular genetic alterations increase the fitness of a cell, causing a clonal expansion.

deleterious mutation

A mutation that decreases the fitness of an organism or cell.

advantageous mutation

A mutation that increases the fitness of an organism or cell, also known as a driver mutation.

epimutation

A non-genetic heritable epigenetic change in DNA methylation or the chromatin.

phenotypic plasticity

The ability of a cell to adopt different phenotypes without changing its genotype.

lineage tree

A type of branching graph diagramming ancestral relationships with only a single path between any two given nodes in the tree.

hybrid cells

Cells that directly derive from more than one ancestor.

vertical inheritance

The passage of properties from parents to offspring.

horizontal gene transfer

The transmission of genetic material from organisms or cells that are not direct ancestors of the recipient.

microvesicles

Lipid bilayer-delimited particles that are released from the cell membrane.

trogocytosis

Transfer of plasma membrane fragments from one cell to another.

reticulated evolution

Transmission of heritable properties from one lineage to another.

niche construction

Modification of the environment by an organism, generally to the benefit of that organism.

transformation

The change from a normal cell state to an malignant state.

neoplastic progression

The process of change from normal tissue to cancer and on to metastatic disease.