4.1 Mechanisms
Once having acquired its independence, the tumor may progress or not by successive additional steps to acquire all the hallmarks of cancers and to develop the competitiveness of its cells [
116]. It could also acquire all these properties at once in a few cells after a catastrophic genome rearrangement (
e.g., chromothripsis, kataegis …). We shall mainly consider the progressive model.
“Tumors evolve by a reiterative process of clonal expansion, genetic diversification and clonal selection within the adaptive landscapes of tissue ecosystems. The dynamics are complex with highly variable patterns of genetic diversity and resulting multiclonal architecture” [
117‐
120]. The process is favored by the mutator phenotype of cancer cells [
121], with the mutation rate increasing in more advanced stages [
18].
The role of the DNA damage response as an anticancer barrier in early tumorigenesis suggests that new genomic instability is an early phenomenon after independence [
122] and contributes to evolution. Among the genetic events, some like those affecting genes of the splicing machinery are frequent and may influence entire programs [
123]. The mutations may also involve recurrent hemizygous, sometimes extensive, deletions [
124]. The reprogramming of characteristics is therefore largely a result of the reprogramming of the genome itself with the emergence of new genomic systems [
125]. Clonal evolution of tumors (
e.g., in leukemia) in response to treatment has been demonstrated by whole genome sequencing [
126].
This evolution also involves epigenetic and possibly
lysogenic-
like and transcriptional control mechanisms [
40,
127,
128]. In colon cancers, epigenetic events are far more frequent than genetic events [
129], and in lung cancer, they may account for the regression of most differentiation genes [
130]. Transcriptional networks, which are able to maintain cellular identity, certainly also have the capacity to play a major role in cancer cell identity [
131].
The Darwinian selection will eliminate many less competitive clones [
132]. The cells will lose physiological characteristics (
i.e., differentiation) that are unnecessary for their survival and reproduction, if not all characteristics of the cell of origin, some of which being used in the new context. The end result is a great diversity and heterogeneity [
133,
134]. Examples are the greatly increased mRNA heterogeneity in cancers [
38] and the great diversity of single-nucleotide mutations from cell to cell in a cancer [
135‐
137]. Cancerization is not a
linear process either in space or in time.
Just the fact that cancer cells proliferate, vary epigenetically, and mutate allows them to discover myriads strategies to escape attacks by anticancer drugs [
138,
139]. Noise increases the adaptability in the cancer cell population. Similarly, the faster increasing entropy of cancer cells,
e.g., in chromatin replication timing [
140], favors the appearance of new mutations, allowing some cells to resist therapies based on existing oncogenic mutations [
141]. For example, treatment of B-Raf mutated melanoma selects cells with Ras mutations which will short-circuit the inhibition and re-establish the activation of MAP kinases in the tumor and also in normal cells, thus generating new skin carcinoma [
142]. Treatment of cells of an osteosarcoma cell line by nutlin which blocks p53 degradation leads to somatic inactivating mutation of p53 and selects for p53 mutated cells [
143]. Inhibition or lack of telomerase activity in human cancers leads to alternative lengthening of telomeres mechanisms [
144]. An analogy between the evolution of drug resistance in bacterial communities and malignant tissue has been suggested [
145].
On the other hand, some mutations or epigenetic controls while not being beneficial or detrimental
per se might introduce a new vulnerability to treatment [
146,
147].
Moreover, the process of adaptation by mutation is not necessarily infinite,
e.g., in bacteria diminishing returns of epistasis among beneficial mutations decelerates adaptation [
148] and lead to an equilibrium of maximal efficiency.
The pervasive increasing entropy in tumor cell evolution will, as in the normal cells, generate far more deleterious than favorable genetic and epigenetic events. This will lead to the development of much more uncompetitive than competitive cells. Tubiana had demonstrated this [
149]: Only 10 to 20 % divisions in breast cancer cells contribute to tumor growth. The less competitive cells may divide a few times but not infinitely, redifferentiate, and become senescent, quiescent, or even die. The other, more competitive cells may correspond to the so-called cancer stem cells postulated by the cancer stem cell theory that we called cancer stem–tumor-propagating cells [
132]. These cells, defined by different sets of properties depending on the systems and the authors, need not be a fixed distinct population of cells, but rather one extreme phenotype in an evolving population [
132,
150,
151].
Thus, the same increase in entropy that generated initially the tumor-initiating cell will impair many cells during the evolution of the tumor but allows new adaptations to arise in a minority.
Due to the stochastic nature of genetic events and their relation with cell divisions it is not astonishing that the best prognostic markers of tumor evolution are the Ki67 proliferative index and tumor size [
31,
32]. The more cell divisions, the more cells involved, and the more mutations, a fraction of which will contribute in the affected cells to positive tumor cell evolution.
4.2 Features of the evolution
Besides the properties acquired in the first transition to independence, mobility, autonomy, and escape of tissue constrains, what are the properties progressively acquired by tumor cells in their evolution to cancer cells, the hallmarks of cancer [
84]?
These properties include the capacity to react independently to the environment (
e.g., O
2 availability, signals from other cells), loss of differentiation, the search for O
2 and substrates; movement, propagation, and exploration behavior; proliferation whenever, wherever it is possible; escape harmful conditions and environment; escape immune reactions; escape of metabolic control; and escape of apoptosis. These correspond to the imperatives of the cancer cell progression and invasion. At the step of invasion, the tumor has acquired the full set of cancer properties, and it is a required step for metastasis. Tumor cells may migrate in bulk as clusters, strands, or sheets (collective migration) or individually. In the latter case, they may, after EMT, either proceed through the extracellular matrix by hydrolyzing and remodeling it or may
swim around it as an amoeba [
152]. These patterns are reversible depending on the context (
e.g., extracellular matrix organization).
The newly acquired properties are not independent; they all reflect the new character of the cell: its cancerishness. A nice illustration of this is the finding that all proposed specific gene expression signatures of breast cancers are related and therefore give redundant prognostic results [
32]. The programs and their resulting hallmarks represent a complex interrelated dynamic network in which some hallmarks themselves drive new programs. In this regard, the
Warburg reorganization of cell metabolism in cancer is both a consequence and a cause of the cancer process itself [
153‐
155]. The Warburg effect confers resistance to anoikis [
92]. The invasion programs imply a close and reciprocal interrelation with the extracellular matrix and the various micro-environmental cancer and non-cancer cells. They present a great variety. The dynamic interrelations of the hallmarks are also illustrated by the many mechanisms of immune escape. Cancer vaccination drives Nanog-dependent evolution of tumor cells toward an immune-resistant and stem-like phenotype [
156]. Hypoxia induces escape from innate immunity through HIF and Adam10 [
157] and promotes tolerance via Treg cells [
158]. Lactate as such stimulates migration, radioresistance, and immune escape [
159]. A large part of the program of embryonic stem cells taken over by cancer cells is the prevention of differentiation.
The evolution of the independent tumor to cancer will involve the acquisition, progressively (evolution) or abruptly (
e.g., after chromothripsis), of all those properties, not necessarily in a given sequence. At any step, the progression of a tumor clone may stop, but this will have little general effects at stages when many clones coexist. Also, higher saturation density is increasingly permissive for expression of the neoplastic transformation [
94].
This evolution will involve the acquisition of all these properties leading to tumor growth, the required angiogenesis, invasion, and ultimately metastasis. Exaggerated angiogenesis will increase but also restrain tumor access and become a pathological process
per se [
160]. The evolution will result from branched genetic evolution within the tumor [
134] as well as even more frequent epigenetic changes [
161‐
163].
For this evolution, the cells will transiently or permanently appropriate programs that improve their competitiveness and growth. The over-expression by human cancers of genes that are specific to a variety of normal human tissues [
164] may reflect this adaptation or/and the increasing entropy of cancer cells. Most often, cancer cells do not invent new programs; they co-opt or highjack available programs expressed in some circumstances by certain cells such as, for invasion, the wound program [
165]; for growth, the regrowth program of amputated liver; for adaptation, to flux the program of blood cells; for escape of differentiation constrains, programs of embryonic stem cells, etc. One other example is the appropriation of leukocytic trafficking and leukocyte attraction mechanisms for invasion and metastasis [
166]. To provide itself with the necessary building blocks for proliferation, the cancer cell activates permanently a temporary program of the normal cell cycle [
167], thus generating a permanent Warburg effect [
168].
Versatility in the use of such programs confers to the cell a nimbleness that reminds of the embryonic cells. For instance, cMyc oncogene when expressed induces proliferation but suppresses cancer metastasis [
169],
i.e., negative and positive roles played by over-expressing the same protein.
In a given cancer, there is complementarity, a synergy between amplified and mutated genes in eliciting programs [
170]. Many cells need an active PI3K pathway to grow, and when they have no activating mutation of the pathway, they over-express IGF1 receptors [
162]. When the pathway is inhibited by therapy, recurrences make use of a Met or cMyc amplification [
171]. Similarly, the widespread potential of cancers to adapt to treatment may involve the redundancy of growth factor-regulated signaling cascade [
172], heterodimeric activations [
173], emergence of new mutations [
174], etc.
The respective use of all these programs varies for different cancers and for one cancer at different times and locations, depending on the existing conditions and the cells accessible programs [
175]. For instance, telomere dysfunction in the absence of telomerase generates very disadvantageous genomic instability but is often corrected by re-expression of telomerase [
126]. EMT transition can suppress major attributes of human epithelial tumor-initiating cells [
176]. Plasticity, reciprocity, and evolution are the major characteristics of these processes [
175,
177].
The evolution of cancer toward independent and maximal growth, and loss of differentiation characteristics leads to a convergence of phenotypes between cancers of same type with very different genotypes [
178] and even between different cancer types [
178,
179], especially for the most dedifferentiated cancers. However, the programs used will often still reflect characteristics of the cell of origin, which are useful for their progression. For instance, transcription factor TTF1 (also called NKX2.1), which is used and necessary in thyroid and lung normal cells proliferation is still used and necessary for pulmonary and thyroid cancer cell lines growth [
180].
Not all cancers use all available programs. For instance, basocellular skin carcinoma or brain tumors may not express EMT program nor develop metastasis. Also, not all the cancer cells in a cancer express the same programs, nor do they necessarily express them all the time,
e.g., EMT is induced in some peripheral carcinoma cells and is reversible after metastasis [
181].
A step of the evolutionary process is the development of the cancer independently of its initial cause,
e.g., the skin tumor in response to a mutagenic then progression treatment, which develops independently of the causal oncogenic events thereafter. By contrast, a hyperplastic tumor may develop with a hormonal treatment or with the expression of an oncogene, both of which recede when the stimulus is stopped [
182]. In the first case, the tumor cells have acquired an independent property, but, in the other, did not. Oncogenic addiction, the necessary activity of one oncogene for tumor progression characterizes this latter type. Because of its mechanistic simplicity and economy, it is very efficient but offers simple therapeutic targets. The inverse, oncogenic independence may arise following treatment (
e.g., by relief of a negative feedback on EGFR in Raf-induced melanoma) [
183,
184]. It explains some resistances to single target treatments. Non-oncogene addiction may also occur, for example, to heat-shock factors that compensate for protein denaturing stress resulting from the increasing entropy of cancer cells [
124,
185].
The classical scenario of tumor progression to cancer described so far should not obscure the fact that many tumors do not progress or even regress. This is a well-recognized possible fate of benign tumors such as hemangiomas, lymphangiomas, gliomas, etc. Thyroid microcarcinomas, with all the characteristic of classical papillary carcinomas (
e.g., B Raf mutations, RET-PTC rearrangements), are found in up to 30 % systematic autopsies. Most never evolve [
186]. Similarly, naevi may be considered as dormant quiescent melanoma [
187]. There are many dormant breast carcinomas [
188], prostatic intraepithelial neoplasia [
189]. Mechanisms of these evolutions are poorly reversible senescence [
189] which may result from the uncorrected oncogenic stimulus itself [
189] or the control by T cell infiltration [
190].