Is cancer the result of uncontrolled cellular growth? a glance into the tumorigenic process.

Cancer cells show metabolic alterations that seem to give them physiological advantages over normal cells. When it comes to generating ATP as the “energy coin”, cancer cells metabolize glucose by glycolysis, even under normal oxygen levels, rather than drawing upon oxidative phosphorylation, a much more efficient mechanism (Warburg effect) [ 73 ]. Indeed, aerobic glycolysis is recognized as a hallmark of cancer. Aberrant metabolism of neoplastic cells results from oncogenic activation of certain signaling pathways (e.g., several genes that are activated by Hypoxia-Inducible Factor decrease the dependence of the cell on oxygen, whereas factors such as Myc, Ras, and Akt upregulate glucose consumption and glycolysis). This abnormal metabolism is selected by the tumor microenvironment, which contains low oxygen and low pH regions [ 74 ]. Metabolic reprogramming, however, goes beyond glycolysis. Evidence exists indicating that cancer cells show metabolic plasticity that includes increased mitochondria biogenesis and activity. Indeed, cancer cells show increased anaplerotic supply of intermediary metabolites for the synthesis of fatty acids, nucleotides, amino acids, cholesterol, glucose, and heme [ 75 ]. Furthermore, it seems that although cancer cells maintain a high glycolytic rate, a significant proportion of ATP production still derives from oxidative phosphorylation, being adenylate kinase a key player [ 76 ]. The altered metabolism gives cancer cells survival and proliferative advantages, such as increased biosynthesis of macromolecules, increased paracrine and autocrine signaling, and evasion of apoptosis [ 74 ]. Interestingly, metabolic needs and vulnerabilities seem to be tumor-specific. Furthermore, such metabolic phenotypes change as tumor progresses [ 77 ]. Indeed, metabolic requirements are different at tumor onset and during metastasis. For example, in cancer of the lung, early mutations in KRAS induce accumulation of D-2-hydroxyglutarate in IDH-mutated tumors, and this, in turn, may favor tumor initiation, driving nutrient uptake and anabolism. STK11 mutations increase tumor aggressiveness, but also induce sensitivity to some metabolic inhibitors. Late KEAP1 mutations, in turn, enhance resistance to oxidative stress, but also glutaminase dependence to supply precursors to produce glutathione [ 78 ]. Meanwhile, in the metastatic cascade, mechanisms such as extracellular acidification that favor intravasation, resistance to oxidative stress in circulating tumor cells, a permissive nutrient environment in micrometastases, and anabolic metabolism in macrometastases, are activated [ 77 ]. These evidences are indicative that metabolic phenotypes evolve and adapt in parallel to tumor progression; thus, raising areas of opportunity for therapeutic approaches targeting metabolic changes and metastatic properties. In normal conditions, cell cycle progresses under the influence of a wide variety of external and internal signals, from receptor molecules on the cell membrane to transcription factors located within the cell nucleus [ 79 ]. Four different cell cycle phases (G1, S, G2, M) have been identified, which are under the control of several cyclins, cyclin-dependent kinases (cdk), and cdk inhibitors. Particular elements controlling the transition between cell cycle phases (e.g. G1-S; G2-M) have been identified (cell cycle checkpoints). A further checkpoint has been identified in the M phase (spindle checkpoint). In the majority of tumors, proteins regulating the G1-S transition are inactivated. G2-M controlling proteins are altered less frequently. Mutations in the p53 gene -which acts as a tumor suppressor transcription factor- are the most common cause for the loss of function of proteins regulating the G1-S transition [ 80 ]. Point mutations and deletions of the Retinoblastoma gene -another key participant in the G1-S transition- are also found in a large number of human cancers [Molinari, 2000]. In normal cells, the G2 phase is arrested when there is DNA damage, so that DNA-repairing proteins can act upon the altered regions of the genome. In cancer cells, however, the cytoplasmic-nuclear signaling pathway that leads to G2 arrest after damage of DNA can be altered due to mutations in specific genes, such as p53 [ 81 ]. It is noteworthy that several cell cycle regulators play key roles at different levels of the cancer cell metabolism. For instance, CycD1 represses β-oxidation and downregulates the synthesis of lipids; CDK1 phosphorylates several subunits of complex I of the electron transport chain, and phosphorylates Drp1, thus controlling mitochondria fission; CDK6 inhibits the activity of PFK1 and PKM2; CDK4 phosphorylates FLCN which mediates mTORC1 recruitment to the lysosomal surface; CDK12 contributes to the phosphorylation of 4E-BP1 promoting translation of mTORC1 target genes [ 82 ]. The potential therapeutic implications of these findings are obvious. Table 3 shows some of the cell cycle regulators that are most commonly altered in cancer. Table 3 Representative cell cycle genes commonly altered in human cancers* Gene Protein Function Alteration in cancer CCNE1 Cyclin E1 Positive regulator of CDK2 Overexpressed CCND1,2,3 D cyclins Positive regulators of CDK4/6 Overexpressed CDKN2A p16, INK4a CDK4/6 inhibitor Mutated, deleted, methylated CDKN1B p27 CDK2 inhibitor Underexpressed RB1 pRb Represses E2F transcription Mutated, deleted AURKA Aurora A Mitotic kinase Overexpressed TP53 p53 Checkpoints, apoptosis Mutated, deleted MTBP MDM2 p53 inhibitor Overexpressed CDKN2A p14 p53 activator Mutated, deleted ATM ATM Checkpoints, repair Mutated, deleted *More than 40 genes have been implicated in neoplastic cell cycle; this table includes only 10 of them. Modified from Table 6.1 in [ 71 ] Representative cell cycle genes commonly altered in human cancers* *More than 40 genes have been implicated in neoplastic cell cycle; this table includes only 10 of them. Modified from Table 6.1 in [ 71 ] Cell growth is a broad term that may refer to the increase in cell size (mass accumulation) or the increase in cell number (proliferation). Although both processes are clearly separable, they are commonly used interchangeably. Cell size and cell proliferation are tightly linked to the cell cycle and its regulation [ 83 , 84 ]. Cell size is controlled by the cell’s internal (genetic) program and by external signals provided by the microenvironment in which the cell develops. Such signals include mitogens, growth factors, nutrients and mechanical signals [ 85 ]. Some internal signaling pathways and specific molecules have been identified as key regulators of cell size; Myc, mTOR, and Hippo are among the most important ones [ 85 – 87 ]. In mammals, most of the cell size increment occurs during the G1 phase of the cell cycle [ 88 , 89 ]. Cell proliferation, on the other hand, refers to the increase in cell number through cell division (one “mother” cell gives rise to two “daughter” cells). This process is also regulated by intrinsic (cell cycle elements, transcription factors, cytokine receptors, signaling pathways, telomeres, etc.) and extrinsic (microenvironmental cells, cytokines, hormones, chemokines, extracellular matrix, etc.) factors. It is important to point out that there is a complex and finely-tuned cross talk between mechanisms governing cell size, proliferation, metabolism, biosynthesis, apoptosis, differentiation, etc. in order to maintain cellular homeostasis [ 90 ]. Increased cell proliferation represents one of the primary hallmarks of cancer, whereas increased cell size (arising from enhanced metabolic activity, such as elevated transcriptional activity, and changes in cell content, including DNA polyploidy, copy number variation (CNV), and C value, or differentiation) may not always be altered in cancer cells. Based on the measurement of different cancer cell lines, as well as normal circulating leukocytes, lymph node cells, and splenocytes, it was observed that, although there is great heterogeneity, tumor cells are usually larger than normal cells [ 91 ]. The relationship between cell size and cancer involves biological factors beyond mere morphological characteristics. Studies by Jones et al., show that melanoma cells with BRAF mutations were small, whereas K-RAS-mutated cells were the largest. CCND1 protein levels may contribute to these differences in cell size and proliferation [ 92 ]. The authors suggest that smaller tumor cells are more vulnerable to DNA-damage targeting agents, such as chemotherapy combined with targeted therapy. Cells’ larger size may make them more responsive to immunotherapy. Also observed in melanoma that larger cells can exhibit signs of senescence even while continuing to proliferate. Based on analysis of more than 7000 tumors, both spatial transcriptomic and microscopic images, a tool was developed that estimates the transcriptional profile of tumor cell populations from microscopic images and contributes to the stratifications of risk and recurrence [ 93 ]. Similarly, when assessing the size of tumor circulating cells from patients with breast, prostate, ovarian, or colorectal cancer, tumor cells showed larger sizes, as compared to normal cells [ 94 ]. Interestingly, it has been found that within the great variability observed among tumor cells, those with small size -which usually include the so-called cancer stem cells- possess higher proliferation and self-renewal capacities and show a more aggressive behavior [ 95 ]. As mentioned before, in normal cells, cell size and cell division are coupled, so that cell homeostasis is maintained. In contrast, in cancer cells, the cell cycle is altered, and so is the cell division process (proliferation); thus, cancer cells are not only very heterogeneous in size but tend to increase their numbers to abnormal levels. In normal cells, programmed cell death (PCD) is actively regulated (activated or repressed) by specific genes and pathways. Different forms of PCD have been described, including apoptosis and autophagy. Apoptosis is driven by a caspase cascade, together with the participation of both proapoptotic and antiapoptotic factors, present at specific concentrations; in fact, it is not their absolute quantities, but the ratio at which they are present which determines if PCD proceeds or not. Apoptosis is characterized by morphological changes, such as cell shrinkage, blebbing of the plasma membrane, mitochondrial fragmentation, and chromatin condensation, without eliciting inflammatory responses. This type of PCD plays essential roles during development and is necessary for tissue homeostasis. This process has been further classified into extrinsic (initiated by activation of death receptors -Fas and TNF receptor- on the plasma membrane) and intrinsic (resulting from cellular alterations such as DNA damage or severe oxidative stress) apoptosis [ 96 ]. A hallmark of cancer cells is their ability to evade apoptosis. Indeed, reduced apoptosis or resistance to it play a key role in carcinogenesis. Apoptosis evasion in cancer involves three mechanisms: disruption of the balance between proapoptotic and antiapoptotic proteins, inhibition of caspase activity, and impairment of cell death receptor signaling pathways [ 97 ]. Interestingly, every abnormality or defect in the apoptotic pathway may be a specific target for cancer treatment. Autophagy is a catabolic process in which lysosomes degrade and recycle damaged, aged or excess biological macromolecules and organelles. It is essential for cellular homeostasis under normal physiological situations, and functions as a mechanism for cell preservation under conditions of nutrient deprivation, hypoxia, or endoplasmic reticulum stress [ 96 ]. Unlike other forms of PCD, autophagy is regarded as a cell-protective mechanism. Its role in cancer is a two-fold: on the one hand, it acts as a tumor suppressor by removing defective organelles, reducing oxidative stress, and preventing DNA damage [ 98 ]. On the other hand, in advanced tumor stages, it promotes tumor progression by meeting the energy demands of tumor cells under the pressure of the tumor microenvironment and mitigating cytotoxicity [ 99 ]. Targeting PCD as a therapeutic strategy in cancer is not only feasible, but is an approach that has already been explored with promising results [ 97 ]. Metabolic, cell cycle, and PCD alterations in cancer cells have important implications in their physiology, favoring a variety of processes such as continual proliferation. Indeed, cancer cells have the ability to sustain chronic proliferation, regardless of the presence of inhibitory signals. Mitogenic signaling in cancer cells may include autocrine (cancer cells produce and use their own mitogenic factors, such as cytokines and growth factors) and paracrine (cancer cells stimulate cellular elements of the tumor microenvironment, which, in turn, reciprocate by producing proliferative elements that stimulate cancer cells) mechanisms. Tumor cells can also display elevated numbers of cytokine-receptors on their surface; thus, they become hyperreactive to stimulatory cytokines. In some cases, intracellular signaling pathways become constitutively activated, so that, there is no need for the interaction between receptors and their ligands to keep cell proliferation going [ 6 ]. The altered physiology of tumor cells has several other implications. Cell-to-cell contact-induced growth inhibition, a mechanism necessary for the generation of tissues, is altered in cancer due to alterations in the expression/action of factors such as Merlin and LKB1 [ 100 , 101 ]. Secretion of cytokines, such as TGF-β, and chemokines, such as CCL21, by neoplastic cells has been identified as a mechanism by which tumor cells evade immune destruction [ 102 , 103 ]. Other features of cancer cells, such as therapy resistance [ 104 ], increased angiogenesis [ 105 ], and tissue invasion and metastasis [ 106 ], are also the result of an altered cell physiology that resulted from genetic/epigenetic alterations.

As mentioned above, cancer begins with the transformation of a normal cell into a neoplastic cell, a process driven by genomic and epigenomic alterations, that may include chromosomal abnormalities (Fig. 3 ). Based on this concept, cancer has been regarded as a monoclonal disease; that is to say, it originates in a single cell from which neoplastic growth progresses [ 107 ]. The concept of cancer monoclonality was initially proposed based on cytogenetic and biochemical evidence [ 108 ]; more recent approaches, including next-generation sequencing, have confirmed this notion [ 109 ]. However, accumulating data also support a polyclonal origin, highlighting the complexity and heterogeneity of tumor evolution. Neoplastic transformation will progress only when the transformed cell is a stem cell, or when the cell acquires stem cell characteristics upon transformation. Stem cells (SCs) are primitive, undifferentiated, and unspecialized cells involved in maintaining the integrity of a particular tissue throughout lifespan. All SCs possess two fundamental characteristics: self-renewal ability and multilineage differentiation potential [ 110 ]. Self-renewal is defined as the capacity of a SC to give rise to new SCs through two possible cell division mechanisms: symmetrical (originating two identical stem cells with self-renewal potential) or asymmetrical (originating a self-renewing stem cell and a non-self-renewing cell with a certain degree of differentiation). Multilineage differentiation potential refers to the capacity of a SC to generate cells of different linages with specific morphologies and functions. Between SCs and mature cells different cell populations are generated which are referred to as transit-amplifying cells, and include progenitor and precursor cells [ 111 ]. SCs represent less than 1% of the total cells in its tissue of origin, and cannot be recognized by their morphology, so they must be identified by their immunophenotype and/or by functional assays, in vitro and/or in vivo [ 110 ]. Additionally, SCs are located in very specific regions within a tissue, known as niches, that are able to regulate stemness and tissue homeostasis [ 112 ]. SCs sit at the top of the lineage hierarchy of any given tissue. In 1994, Dick`s group demonstrated that, similar to normal hematopoiesis, in Acute Myeloid Leukemia (AML) there is a cell population with SC characteristics that is able to recapitulate the disease when injected into immunodeficient mice [ 113 ]. The cells within this population were called Leukemic Stem Cells (LSCs). LSCs are located in the apex of the leukemic hierarchy, and are regulated by mechanisms similar to those regulating normal hematopoietic SCs. Few years later, the existence of SCs in breast cancer was reported by Clarke and co-workers [ 114 ]. The work by Dick’s and Clark’s groups demonstrated that only cancer cells with SC characteristics were able to initiate and promote neoplastic development. Since then, the existence of Cancer Stem Cells (CSCs) has been reported in practically all types of cancer, and although its proportion is variable -ranging from 0.2% to 82.5%- its frequency has been associated with tumor progression (Fig. 3 ) [ 115 – 117 ]. Fig. 3 Cancer stem cells and their niche. Most cancers originate from a rare population of cells that are able to both self-renew and give rise to the rest of the tumor cells. Such cancer stem cells (CSCs) are present at low frequencies and reside within the tumor microenvironment in specific sites known as niches. CSCs interact with the different cellular and molecular elements that constitute the tumor mass, including fibroblasts, macrophages, lymphocytes, endotelial cells, and the rest of the tumor cells, as well as extracellular matrix, cytokines, chemokines, ions, etc. A fraction of the CSCs are quiescent (G0 phase of the cell cycle); thus, they are not affected by radiation or chemotherapy. CSCs present genetic and epigenetic alterations that impact in their metabolism, intracellular signaling, cell surface protein expression, and specific mechanisms to eliminate drugs. Cell surface antigens relevant in LSC phenotype include CD25, CD34, CD44, CD90, CD96, CD117, CD123, and IL1-RAP. Antigens relevant in CSC phenotype include CD24, CD29, CD44, CD49f, CD70, CD133, CXCR-4, LGR5, and EpCAM Cancer stem cells and their niche. Most cancers originate from a rare population of cells that are able to both self-renew and give rise to the rest of the tumor cells. Such cancer stem cells (CSCs) are present at low frequencies and reside within the tumor microenvironment in specific sites known as niches. CSCs interact with the different cellular and molecular elements that constitute the tumor mass, including fibroblasts, macrophages, lymphocytes, endotelial cells, and the rest of the tumor cells, as well as extracellular matrix, cytokines, chemokines, ions, etc. A fraction of the CSCs are quiescent (G0 phase of the cell cycle); thus, they are not affected by radiation or chemotherapy. CSCs present genetic and epigenetic alterations that impact in their metabolism, intracellular signaling, cell surface protein expression, and specific mechanisms to eliminate drugs. Cell surface antigens relevant in LSC phenotype include CD25, CD34, CD44, CD90, CD96, CD117, CD123, and IL1-RAP. Antigens relevant in CSC phenotype include CD24, CD29, CD44, CD49f, CD70, CD133, CXCR-4, LGR5, and EpCAM The isolation of LSCs and CSCs has been based on the expression of different cell markers, many of them shared with the normal analogous tissues. In myeloid leukemias, for example, LSCs express CD34 but not CD38 just as it happens with normal hematopoietic SCs. Nevertheless, a growing number of molecules associated with LSC phenotype has been described, and this include: CD25, CD34, CD44, CD90, CD96, CD117, CD123, IL1-RAP, among others [ 115 ]. On the other hand, the isolation of CSCs in solid tumors has focused on the expression of CD24, CD29, CD44, CD49f, CD70, CD133, CXCR-4, LGR5, EpCAM, etc. (Fig. 3 ), some of them also detected on the surface of normal SCs. Additionally, intracellular markers as transcription factors (OCT4, SOX2, Nanog), cytoplasmic proteins (ALDHs), and other cell proteins (Musashi 1 or 2, Alpha-Fetoprotein, BMI-1 and Dcamkl-1) have been related with CSCs in solid tumors [ 116 ], confirming that tumors -and the CSCs or LSCs present within them represent highly heterogeneous entities with very specific regulatory mechanisms. CSCs and LSCs keep signaling pathways in active state; most of them are involved with proliferation, self-renewal, differentiation, immune regulation, epithelial-mesenchymal transition, apoptosis inhibition, and metastasis. These pathways involve molecules such as MAPK/ERK, JAK/STAT, PI3K/AKT, mTOR, Wnt, Hedgehog, Noth, NF-kB, Hh, Notch, TGFb/Smad, and Bcl6 [ 118 , 119 ]. Additionally, CSCs and LSCs are able to change the way of getting energy through a metabolic reprogramming process, giving them an advantage for survival and proliferation under nutritional deficiency, while also regulating autophagy, immune evasion, and their interaction with several elements of the tumor microenvironment [ 120 ]. Interestingly, and despite that many signaling pathways in CSCs and LSCs are turn-on most of the time, the existence of a cellular fraction within such cell populations that remains in a G0 cell cycle state has been reported. These quiescent cells, similar to what is found in normal cells, have a small size, lower levels of proteins and RNA, and reduced mitochondrial activity [ 121 ]. G0 CSCs are directly related with therapy-resistant phenotypes, since they are insensitive to most antineoplastic agents that exert their effects on actively cycling cells. Cell cycle quiescence, together with the increased presence of drug efflux pumps, pro-survival signaling pathways, DNA repair mechanisms, and protection by microenvironment components suggest a strict physiological control that allows for the permanence of CSCs/LSCs and this, in turn, favors tumor development, resistance to treatment, dissemination, and relapse. Current evidence indicates that tumors can be viewed as large populations of genetically diverse cells. Indeed, there is great cellular heterogeneity within a tumor, with different cell populations being part of it [ 122 ]. However, if, as discussed previously, cancer is a clonal disease that originates from a single CSC, how is such a heterogeneity achieved? Once the initially transformed cell (CSC) (Fig. 4 a i, ii) starts proliferating (Fig. 4 a iii), its progeny presents with genetic instability, so that, although all the cells belong to the same clone, there is genetic diversification among them [ 123 ]. In this way, the primary tumor is initially dominated by a single, yet genetically diverse, clone (Fig. 4 a iv). Although genetic instability has long been recognized as a feature of cancer cells [ 124 ], the elements that induce it are only recently being identified. It is now known that factors such as the cytosine deaminase APOBEC3B, a potent mutagenesis promoter, and the presence of extrachromosomal DNA, favor genetic instability and chemotherapy resistance [ 125 – 127 ]. Over time, the initial clone is subjected to selective forces, as well as neutral changes, and, eventually, new (sub)clones start to appear, all of them derived from the initial one (Fig. 4 a v). Particular conditions, including endogenous elements, such as the dynamic functional integrity of the tumor microenvironment, anatomical barriers, and the patient’s immune response, as well as exogenous elements, such as therapeutic agents infused into the patient, will eliminate particular subclones and will promote the growth of others, so that the clonal landscape of the tumor will evolve in a Darwinian fashion [ 128 ]. Clonal evolution is a very dynamic process. Indeed, the same tumor mass will comprise different clone scenarios throughout time (Fig. 4 a vi-viii); this means that, at one time, the tumor may be sensible to a particular drug, but, at a different time, the tumor may be refractory to the same drug. Clonal evolution is driven by both, cell intrinsic and cell extrinsic factors. The former include genomic modifications, whereas the latter include external (molecular and/or cellular) forces acting on the tumor cells. Genomic modifications are central to evolution since they provide the genetic diversity that serves as substrate for selection to operate. They may include point mutations (single-base-pair changes), insertions and deletions of base pairs, and larger structural modifications, such as whole genome doubling, chromosomal loss or gains, and translocations [ 129 , 130 ]. Fig. 4 Cellular transformation and clonal evolution in tumor development. (a) The figure illustrates the transformation of a healthy stem cell (SC) into a cancer stem cell (CSC) (i-ii) and how the resulting cell populations evolve over time through a dynamic process of clonal selection, influenced by intrinsic and extrinsic factors (iii-v). Selective pressures, such as therapeutic agents or immunological barriers, promote or eliminate subclones, altering the tumor’s clonal landscape. This process enables the emergence of subpopulations with adaptive characteristics, which can lead to therapeutic resistance and tumor progression (vi-viii). (b) Diagram representing, in evolutionary terms, the trunk and branched mutations that significantly contribute to the generation of subclones Cellular transformation and clonal evolution in tumor development. (a) The figure illustrates the transformation of a healthy stem cell (SC) into a cancer stem cell (CSC) (i-ii) and how the resulting cell populations evolve over time through a dynamic process of clonal selection, influenced by intrinsic and extrinsic factors (iii-v). Selective pressures, such as therapeutic agents or immunological barriers, promote or eliminate subclones, altering the tumor’s clonal landscape. This process enables the emergence of subpopulations with adaptive characteristics, which can lead to therapeutic resistance and tumor progression (vi-viii). (b) Diagram representing, in evolutionary terms, the trunk and branched mutations that significantly contribute to the generation of subclones In terms of cancer evolution, two types of genomic modifications (mutations) can be identified: truncal mutations, those appearing early in time, driving tumor initiation and expansion, and branched mutations, those appearing at later time points and associated with the origin and development of different subclones (Fig. 4 b). One of the best examples of truncal mutations would be the reciprocal translocation giving rise to the BCR-ABL fusion gene, responsible for the development of chronic myeloid leukemia [ 131 ]. Interestingly, it has been recognized that non-standard Darwinian processes may also contribute to the evolution of tumors beyond somatic selection [ 132 ]. For example, it has been generally assumed that mutations appear sequentially, in a gradual manner, over time; however, recent evidence indicate that a large number of genomic alterations may appear in very short periods of time in cancer cells [ 133 ]. It has also been observed that cancer-driving genomic alterations may be selected for and accumulate prior to tumor initiation, as a result of aging or specific chemical insults. Such mutations would be sufficient to induce cancer development, with a variety of cell populations being present from the beginning, so that no further alterations would be necessary for the development of subclones [ 134 ]. Tumor growth and progression cannot be conceived without clone selection, that is, the increase in the frequencies of particular clones within a tumor. The causes and mechanisms of selection, however, can vary widely across space and time. Indeed, clone fitness vary depending on the external forces operating in a particular tumor and at a particular time. Clonal dominance within a tumor can result from positive (one or a few clones have advantages over other clones, so that, their frequencies increase and become dominant within the clonal landscape) or negative (some clones are at a physiological disadvantage and disappear; thus, the remaining clones will dominate the clonal landscape) selection. Increased proliferation is one of the main features of the selected clones, particularly during tumor initiation; however, other hallmarks, such as reduced apoptosis, increased metabolism, immune evasion, and cell motility, can play important roles in favoring the development of particular tumor cell subpopulations (Fig.  4 a iii-vi). Importantly, during oncologic treatment (chemotherapy, radiation, targeted therapy, immunotherapy), new external forces operate, acting directly on the neoplastic cells, and indirectly -by changing the selective pressures imposed by the tumor microenvironment over the neoplastic cells- so that the clonal landscape of the tumor is modified (Fig. 4 a vi-viii). For example, neutral mutations -previously “irrelevant”- may become adaptive inducing the emergence or disappearance of particular clones. Interestingly, it has been found that treatment can induce loss of clonal diversity, and this, in turn, may result in resistance [ 135 ]. Based on the above concepts and evidence, it is clear that intratumor heterogeneity is related to its clonal evolution, genomic/phenotypic diversity, and treatment resistance, which, altogether, form a complex scenario that has important clinical implications [ 25 ]. As indicated above, most evidence presented to date favor the notion that cancer is a monoclonal neoplastic disorder. Cytogenetic, biochemical, and genomic studies support this concept [ 108 , 109 ]. However, evidence from mutational profiles, X-linked marker heterogeneity, CNVs, cytogenetic alterations, chimeric and transgenic mouse models, and single-cell sequencing, suggest a polyclonal origin in certain types of breast, colorectal, esophageal, and prostate cancer [ 136 ]. While monoclonality implies that all tumor cells share early driver mutations, polyclonality suggests that such alterations are restricted to particular cell subpopulations. A polyclonal origin of cancer would imply that two or more tumor initiating cells develop simultaneously, most likely due to the effect of the same carcinogen, and each one of them follows its own evolutionary trend [ 137 ]. Whereas the monoclonal model suggests that clonal diversity is gradually achieved as a result of divergence of new clones throughout time, the polyclonal model suggests that several clones are present in the early stages of tumor development, and the tumor becomes more homogeneous at later stages due to clonal selection and expansion [ 137 , 138 ]. Discerning between monoclonal vs. polyclonal tumor development in clinical settings is rather difficult due to the fact that, when a tumor is detected, the neoplastic growth has already achieved a relatively advanced progression stage; thus, there is no access to the very early tumorigenic processes. Even when two or more separate neoplastic growths are found within the same tissue, we cannot rule out any one of the two models, since such neoplastic growths could arise from different tumor-initiating cells (polyclonal origin), or from the same single cell that produced progeny with the ability to migrate within the tissue (monoclonal origin). Interestingly, the detection of driver mutations in tissues without neoplasia highlights the need to consider additional factors, such as interclonal cooperation, in tumor development [ 136 ]. Thus, further studies, which will certainly involve genomic and epigenetic analyses, using animal models, are warranted in order to clarify this issue.

From the early days of cancer research, in the late 19th century, and throughout most of the 20th century, a tumor was viewed as a chaotic conglomerate of cells. Indeed, based mainly on its macroscopic appearance and numerous histological studies, a tumor was considered as a random mixture of cells and extracellular matrix [ 139 , 140 ]. During the last three decades, however, this view has changed. Like any organ, a tumor consists of the main cell type -in this case, a heterogeneous population of neoplastic cells- that interact with different types of non-cancerous cells and extracellular matrix (ECM) molecules, which, together, constitute the tumor microenvironment [ 141 ]. The interactions between neoplastic and microenvironment cells are complex, involving a bidirectional cross-talk using a “molecular language” that comprises a vast array of soluble and cell-associated proteins, including cytokines and chemokines [ 31 ]. Tumor microenvironment cells include fibroblasts, epithelial, myoepithelial, adipocytes, endothelial, perivascular, and mesenchymal stromal cells, as well as several types of immune cells, such as macrophages, neutrophils, mast cells, T and B cells, and platelets [ 142 ]. As tumors develop and progress, they undergo significant architectural changes that involve not only neoplastic cells -changes in the frequencies of different tumor subclones- but also different elements of the microenvironment. In this regard, it has been suggested that stage-specific changes of the microenvironment are important in favoring the neoplastic growth. Indeed, it has been indicated that late-stage stroma is more supportive of tumor progression than early-stage stroma [ 142 ]. Elements of the microenvironment, including fibroblasts, type I collagen, and the immune cell infiltrate promote tumor progression in advanced stages [ 143 , 144 ]. As cancer cells proliferate and expand, they produce factors that recruit and activate fibroblasts, which, in turn, produce and secrete a variety of cytokines, such as TGF-β and HGF, and increase ECM production, including type I collagen, and MMPs, which induce ECM modifications [ 145 ]. Changes in ECM content and distribution favor neoplastic progression by architectural remodeling, and these changes in the tumor architecture promote cell invasion by enabling cancer cells to migrate along the fibers of collagen and enhancing integrin signaling [ 144 , 146 ]. Remodeling of the tumor architecture also favors the angiogenic process, which occurs at different stages of tumor progression. Indeed, tumors recruit blood vessels in order to ensure a blood supply that will allow them to obtain enough nutrients and oxygen for their growth. Factors such as VEGFs and FGF2, secreted by cancer cells and by tumor-infiltrating myeloid-derived cells, are key players in the angiogenic process [ 105 ]. Interestingly, there seems to be no correlation between tumor vasculature and tumor aggressiveness. Pancreatic adenocarcinoma, which is very aggressive, is poorly vascularized, whereas the slow growing grade I pilocytic brain tumors, which do not metastasize, are highly vascularized [ 105 ]. This exemplifies that certain factors must be secreted at the right time and in the correct manner to sustain tumor activity, regardless of the tumor’s aggressiveness. It has been suggested that tumor architecture resembles the histological organization observed in early stages of normal tissues. For example, mammary tumors show a multilayered epithelium with poor polarity, similar to the actively proliferating and invading epithelium that is seen in the early stages of the normal developing mammary gland [ 147 ]. It should also be pointed out that the different components of a tumor are not distributed randomly. For instance, ECM deposition and leukocyte infiltration are more pronounced at the tumor-stroma border [ 142 ]. Such histological features are usually taken into account by pathologists at the moment of analyzing and categorizing tumors. Remarkably, specific regions within the tumor microenvironment have been identified that constitute special zones or niches in which cancer cells develop. Similar to the niches found within the hematopoietic microenvironment [ 148 , 149 ], these tumor niches, formed by different stromal cells and ECM molecules, confer particular functions to the cancer cells. It has been proposed that CSCs colonize and establish themselves into specific niches where they self-renew and expand. It is not clear, however, if such CSC niches are formed de novo, or if CSCs hijack preexisting niches that were originally part of the organ and that served for the development of normal stem cells. In any case, CSC niches are key elements in the physiology and development of tumor growth -providing the conditions for CSC expansion and protecting such cells from the effects of chemotherapy and radiation- and may also be important in the development of anticancer therapeutic strategies. Based on the complexity found in terms of their architectural organization and intricate physiology, tumors can be viewed as actual organs which interact not only with surrounding tissues and organs, but with the human body, as a whole.

The epithelial-mesenchymal transition (EMT) is a highly organized and conserved biological process in which epithelial cells acquire properties that allow them to migrate. This implies the transitory loss of certain properties of the epithelium and gaining of properties typical of mesenchymal cells. Epithelial cells form the barrier between the external and internal environment in the organs of the body. They remain together because of intercellular junctions and junctions that bind them to the stroma that supports them, thus their mobility is limited. The changes acquired by cells in EMT include reduction in cell junction properties, loss of apical-basal polarity, and modification of position and types of intermediary filaments that form the cytoskeleton. Consequently, mesenchymal cell characteristics are acquired, such as bipolarity, a long, spindle shape, filipodia, and the capacity to move and interact with the components of the ECM. It is worth noting that this process is reversible; thus, when mesenchymal cells arrive at their destination, the inverse process, known as mesenchymal-epithelial transition (MET), occurs [ 150 , 151 ]. The concept of EMT was initially recognized in embryogenesis by Elizabeth Hay in 1968, and since then has been described in several events in embryonic development such as gastrulation, as well as the formation of the neural crest, cardiac valves, skeletal muscle, and the palate. The inverse process, MET, is also present in stages of early development, such as nephrogenesis [ 151 ]. An additional biological phenomenon in which EMT takes part is scarring and tissue repair, in which fibroblasts arise from epithelial cells that have been damaged, and their adjacent microenvironment has triggered an inflammatory response. Of particular interest in the field of oncology is the epigenetic reprogramming of cancer cells that replicates this process, especially in carcinomas (epithelial malignant neoplasms) that acquire the capacity to invade and metastasize. Based on its various biological contexts, EMT has been classified into three types: Type 1, associated with embryonic development; Type 2, related to scarring; and Type 3, present in cancer [ 152 ]. The complexity of the changes present in EMT requires many molecular factors which, based on the role played in the process, may be divided into effectors, master regulators, and inducers. Many EMT effectors are subcellular, structural proteins that determine epithelial or mesenchymal phenotypes. Regarding epithelial cells, there is an evident reduction or loss of proteins participating in cell junctions, such as E-cadherin, α-catenin, γ-catenin and claudin. E-cadherin, considered the guardian of epithelial phenotype, has a notable role in EMT due to its interaction with other structural proteins, coordinating cell organization and communication with the cell microenvironment, and participating in transcription and cell differentiation [ 153 ]. On the other hand, the loss of E-cadherin function is accompanied by an increase in expression of cadherins typical of mesenchymal cells, such as cadherin 11 and N cadherin. The latter can interact with members of the fibroblastic growth factor (FGF) family that induce a cascade of signals promoting cell invasion and migration [ 154 ]. Intermediary filaments also go through changes in this process, and reduction in keratin filaments, together with an increase in vimentin filaments, more abundant in mesenchymal cells, has been observed together with increased expression of surface proteins participating in cell migration, such as CD44 and integrin β6 [ 155 , 156 ]. The process of EMT implies changes in transcription of many genes that regulate cell adhesion, differentiation, and migration. This requires factors favoring or inhibiting gene expression, depending on when they are needed. Among the transcription factors considered as master regulators of EMT are Snail, Zeb, and Twist, each of which, by a variety of mechanisms that share decreased E-cadherin expression, repress transcription of other proteins present in cell junctions and stimulate proteins typical of mesenchymal cells [ 157 ]. In early stages of embryonic development, Wnt and FCTβ pathways are most frequently used in the process of EMT, although pathways associated with tyrosine kinases, such as epidermal growth factor (EGF) and FGF pathways, also take part [ 151 ]. Thus, it is no surprise that in the context of neoplasms, we find FCTβ, members of the EGF, FGF and HGF families as EMT inducers that generate signals for expression of Snail1, Twist and ZEB. These factors induce epithelial cells to lose adherent junctions, acquire a fusiform phenotype, express enzymes that break down ECM, increase motility and resistance to apoptosis [ 158 ]. Likewise, TGFβ is considered an important EMT promotor as it takes part in regulation of signaling pathways such as Notch, Wnt/β catenin, NFkB, and tyrosine kinase receptors that play a role in regulation and completion of the mesenchymal phenotype in neoplastic cells [ 159 ]. In addition to these factors, the conditions surrounding epithelial cells, such as hypoxia and the cytokines present in inflammation, promote EMT. For example, tumor necrosis factor-α (TNFα) stabilizes Snail1 transcription factor and induces Twist expression, while the response to hypoxia mediated by HIF-1 also favors expression of Twist 1 and Snail1 [ 160 , 161 ]. EMT is not an “all or nothing” phenomenon. Quite the contrary, cell populations undergoing this process are diverse, and a variety of changes can be observed, from a predominantly epithelial phenotype to a mesenchymal one, including populations with hybrid traits that correspond to various intermediary states of differentiation that vary in migration capacity, differentiation, and stem cell markers. Some models have shown that hybrid populations have the greatest plasticity and capacity for propagation, while also expressing the most markers related to stem state. It is also interesting to note that depending on where on the differentiation spectrum cells are found, there are changes to the surrounding microenvironment, and this in turn modifies the number of stromal and inflammatory cells present. Thus, it is no surprise that a hybrid EMT phenotype is found in tumor cells circulating in patients with carcinoma of lung, prostate, breast, liver, colon, stomach and nasopharynx. Similarly, co-expression of epithelial and mesenchymal markers is associated with poor prognosis in these neoplasms [ 162 ]. EMT has morphologic manifestations that can be identified histologically. One of these can be found in carcinomas, in which aggregates of 2 or 4 cells are formed under the influence of Snail1 and Twist, and then break away as buds of the main tumor in the interphase between healthy tissue and the tumor, favoring dissemination. Similarly, some carcinomas, known as sarcomatoid, that acquire a fusiform phenotype like that of mesenchymal cells, express transcription factors associated with EMT, like Snail, Zeb1, Twist and Slug [ 163 ]. In organs such as kidney, bladder, and skin, this phenotype is associated with increased capacity for neoplasm dissemination. The influence of EMT in the biological behavior of cancer is so important that it is currently considered in the molecular classification of stomach, colon, and bladder carcinomas, among others. In colorectal cancer, for example, expression of factors related to EMT in neoplastic cells determines classification into the mesenchymal group, in which neoplasms have poor prognosis due to a greater capacity to invade and metastasize. Considering the substantial cell number and the great cellular heterogeneity within primary tumors, only a few neoplastic cells are able to penetrate the walls of lymphatic or blood vessels, and circulate through the bloodstream. Many of such cells perish while in circulation; however, some of them are able to extravasate and reach other tissues. Once in a distant organ, probably the majority of them die, whereas some remain alive for long periods of time but are unable to induce further lesions. Only a minority of those cells possess characteristics that allow them to form new neoplastic growths, known as secondary tumors. Thus, overall, this process, known as metastasis, can be considered as a highly inefficient process [ 106 ]. It is now clear that metastatic cells are genetically and phenotypically different from the vast majority of the cells present in the primary tumor mass [ 164 ]. Such metastatic cells possess biological features that are found in stem cells; thus, it is now accepted that such cells are actual metastatic stem cells (metSCs) [ 164 ]. In their metastatic journey, metSCs follow a rather complex cascade of events. Indeed, they have to invade adjacent tissues and intravasate into circulation; they have to disseminate in the bloodstream and extravasate in different organs. Once they reach the new microenvironment, they have to survive upon arrival; they may remain latent for varying periods of time (from months to years), and then, they reactivate, proliferate, and generate new tumor masses [ 165 ]. MetSCs may already exist within the primary tumor as part of the intrinsic neoplastic hierarchy described before [ 118 ], or they may derive from non-SCs that are exposed to different signals, including cytokines, such as HGF and TGF [ 166 , 167 ]. It seems that the metastatic traits expressed by metSCs are selected at the invasive front; that is, the intersection of the advancing tumor mass and the surrounding stroma. Such an invasive front is rich in a variety of cell types, including cancer-associated fibroblasts, tumor-associated macrophages, and newly-generated blood vessels [ 168 ]. Also abundant in the invasive front are factors such as TNF-α, TGF-β, Wnt, and Notch [ 169 ]. It is important to point out that not all tumor cells that extravasate and enter the bloodstream are actual metSCs. Only a fraction of those neoplastic cells in circulation will succeed in colonizing distant organs. Circulating tumor cells (CTCs) usually travel as single cells; however, it has been found that tumor cells can also travel as clusters [ 170 ]. In fact, it has been shown that CTC clusters possess higher probabilities to metastasize than single CTCs [ 171 ]. CTC clusters could be homotypic or heterotypic. In the former, tumor cells alone form aggregates in a process favored by molecules such as Plakoglobin (gamma-catenin) and intercellular adhesion molecule (ICAM) [ 170 , 172 ], and conditions such as hypoxia. In the latter, tumor cells form aggregates with neutrophils, myeloid-derived suppressor cells, macrophages, platelets, cancer-associated fibroblasts, or red blood cells. The generation of this type of clusters is favored by cell adhesion molecules such as ICAM-1, VCAM1, and MAC-1 [ 171 ]. For several decades, it was thought that metastasis was the final stage of the tumorigenic process. The traditional view was that genetic and epigenetic alterations accumulate in primary tumor cells over long periods of time, until such cells have matured into a state in which they are ready to spread out. However, in the last 15 years, several studies have presented evidence indicating that tumor cells -and even precancerous cells- can be detected in circulation at early stages of carcinogenesis; thus, suggesting that metastasis may occur during the initial stages of tumor formation, when no evident tumors have been clinically detected [ 165 , 173 – 175 ]. This has been observed in several types of cancer, including breast and pancreatic cancer [ 176 , 177 ]. Based on the complex cascade of events described above, it is clear that metastasis is not a random process. For the most part, it is a selective process in which some stochastic events occur at specific steps of the molecular/cellular cascade (for instance, during the initial cell shedding from the primary tumor, which occur at varying times). Patterns of directionality have been identified in several types of tumors. The dissemination of CTCs can initially occur in a stochastic manner. However, the likelihood of these cells colonizing specific organs is significantly influenced by their intrinsic molecular characteristics and the composition of the lodging tissue, rather than being dictated solely by anatomical location or blood flow. Within metastatic subclones, various gene regulatory and epigenetic programs are activated, shaping the expression of specific adhesion molecules, receptors, integrins [ 178 ], exosomes [ 179 ], and RNAs. This molecular framework directs organ colonization in a non-random fashion and may facilitate the formation of pre-metastatic niches [ 180 ]. Emerging experimental evidence indicates that the metastatic spread of breast cancer to particular organs is guided by well-defined multigenic programs, particularly in the context of bone metastasis. In functional in vivo models, the strategic combination of overexpressing IL11 and OPN, along with CXCR4 or CTGF, endows tumor cells with a highly metastatic phenotype, akin to that of inherently aggressive cell populations that naturally express a comprehensive range of genes associated with bone colonization [ 181 ]. Moreover, studies have demonstrated that subpopulations of tumor cells derived from the same breast tumor can exhibit differing capacities for metastasis to the bone, lung, or adrenal glands. Transcriptomic analyses have unveiled organ-specific metastatic gene sets that operate independently of general poor-prognosis signatures [ 182 ]. Furthermore, breast cancer subtypes vary not only in their invasive potential, measured by the number of invading cells, but also in their tropism for metastatic sites. For example, MCF-7 cells exhibit increased invasion and a preference for bone metastasis, while SKBR3 cells are more prone to disseminate to the liver and lungs. In contrast, MDA-MB-231 cells demonstrate high invasive capability without a specific organ preference. Notably, organoids derived from a patient with lung metastasis have been shown to selectively invade lung tissue in experimental models, while avoiding colonization of the liver or bone [ 183 ]. Further instances include lung cancer, skin melanoma, and breast cancer usually relapse in organs such as bones, lungs, liver, and brain. Prostate carcinoma, on the other hand, relapses in bones, whereas sarcomas in the lungs [ 184 , 185 ], and colorectal cancer in the liver [ 31 ]. Such a directionality is influenced by the circulation patterns and by the active participation of a variety of factors, such as paracrine molecules (e.g., VEGF and EGF), proteases (MMPs and cathepsin), cancer cell-autonomous functions (e.g., invadopodia formation), recruitment of stroma elements (e.g., tumor-associated macrophages), and interaction with blood platelets [ 164 , 186 – 188 ]. Survival and expansion of metSCs is supported by specific sites within the colonized organ called metastatic niches [ 164 ], which consist of particular stromal cells, ECM proteins, and diffusible signals. Metastatic niches play important, active roles in allowing metSCs to develop into secondary tumors. Such niches can be preexisting stem cell niches in the particular tissue colonized by metSCs, or they can be unspecific sites that become metSCs niches due to the conditioning action (secretion of factors that act in a paracrine way) of metSCs themselves [ 189 ]. Factors such as the presence and density of mesenchymal cells, and the expression and deposition of extracellular matrix proteins in the lodging tissue seem to play key roles in the metastatic activity of breast cancer cells into the brain. TNBC and HER2 + subtypes both tend to metastasize to the brain more than to other anatomical sites. However, each subtype establishes metastasis via distinct mechanisms. TNBCs form perivascular structures that interact with astrocytes and other glial cells, creating an infiltrative and inflammatory tumor microenvironment. In contrast, cells from HER2 + tumors form compact, spherical colonies via Tenascin C, resulting in a segregated and phagocytic tumor-microenvironment interface. Although they share common mechanisms, such as a microglial response similar to that observed in Alzheimer’s disease [ 190 ]. Interestingly, not all secondary tumors develop following similar patterns and architectural arrangements; instead, different patterns have been described, some of them forming a dense globular lesions [ 191 , 192 ]. It is clear that certain cancers, such as liver and testicular, present few metastatic sites (2–4). Others, such as melanoma and breast cancer, present between 4 and 8 sites [ 193 ]. Building on these observations, clinical data can be used to identify metastatic patterns. Statistical analysis of clinical case data can be used to estimate the probability of metastasis from a given primary tumor to a specific site [ 194 ]. These clinical observations, along with biological characteristics, will enable improved identification of the conditions that favor metastatic sites and timing, as well as the most effective preventive and curative therapeutic options. Many different questions regarding the biology of metastasis remain unanswered. The identity and origin of metSCs, the patterns of directionality, the understanding of the epigenetic programs that allow metSCs to adapt to new microenvironments, and the mechanisms involved in the development of secondary tumors are among the issues that need to be addressed. Interestingly, the different steps involved in metastasis have now been identified as potential targets for cancer therapies [ 195 ].

Cancer results from alterations in the sequence/expression of certain genes; thus, cancer is a genetic disease [ 7 ]. Cancer is also the result of changes in non-coding segments of the human genome, and changes in the structure/architecture of the chromatin; thus, cancer is an epigenetic disease [ 8 ]. It is generally accepted that a single change is not sufficient for cancer to develop; rather, cancer originates and progresses only when several genetic and/or epigenetic alterations are present simultaneously. What triggers such genetic/epigenetic alterations? A vast number of agents and factors have been implicated, including viral agents, such as the human papillomavirus (HPV) [ 9 ]; bacteria, such as Helicobacter pylori [ 10 ]; chemicals, such as nitrosamines; and radiation, including X-rays and UV light [ 11 ]. It is noteworthy, however, that cancer development does not occur as a result of any fortuitous genetic/epigenetic alteration in any sequence of the genome; in fact, cells continuously suffer genetic and epigenetic alterations that do not lead to cancer. In order to be carcinogenic, a molecular change must occur in specific protein-coding or non-coding sites. Interestingly, certain regions within the human genome seem to be more susceptible to suffer those alterations [ 12 , 13 ]. Furthermore, many genetic/epigenetic alterations seem to be tumor-specific.

We now know that cancer is the result of a sequential series of alterations in well-defined genes that alter the function of a limited number of molecular pathways [ 14 ]. Traditionally, two types of genes have been implicated in the genesis of cancer: oncogenes and tumor suppressor genes [ 7 , 15 ]. Some examples of these genes are presented in Fig. 1 a; Table 1 . Interestingly, some cancer genes may act as oncogenes or tumor suppressor genes in particular tumors (Fig. 1 a). The former consist of a family of genes that, when activated due to mutations, favor tumorigenesis. Activation of oncogenes can result from intragenic mutations that affect specific, crucial nucleotides that regulate the activity of the gene, from gene translocations, or from gene amplification [ 16 ]. Oncogenes include growth factor or tyrosine kinases receptors, genes that encode for cytokines, cytokine-receptors, or proteins involved in intracellular signal transduction pathways, from the plasma membrane to the nucleus. Tumor suppressor genes, on the other hand, are genes that, when inactivated, contribute to tumor growth [ 15 , 17 ]. Their inactivation can occur by missense mutations at nucleotides that are essential for the gene activity, from mutations that result in a truncated protein, from insertions or deletions, or from epigenetic silencing (e.g., DNA methylation or the action of non-coding RNAs). A third type of cancer genes comprises the stability genes, which are genes whose proteins participate in keeping genetic alterations to a minimum. Stability gene products include DNA repair proteins (those repairing small alterations in DNA, such as mismatch repair, nucleotide-excision repair, and base-excision repair proteins) and proteins that control processes involving large portions of chromosomes, such as those controlling mitotic recombination and chromosome segregation. Their absence or alteration significantly increases genome mutations [ 18 ]. Cancer genes (oncogenes or tumor suppressor genes) that, when altered (due to mutations or epigenetic changes), directly contribute to cancer development and progression are known as driver genes. Changes in driver genes provide a selective advantage to the cell, allowing the formation of a malignant tumor. Figure  1 b-c illustrates an example of driver genes identified in head and neck cancer and breast cancer. Furthermore, some genes may act as oncogenes in certain types of cancer and as tumor suppressor genes in other types. The organized development of cancer requires, paradoxically, a disruption in DNA repair molecules. However, this disruption follows specific patterns that allow the selection of cells with proliferative advantages. Thus, the progressive inactivation of repair molecules contributes to a structured evolution of cancer. Fig. 1 Key oncogenes, tumor suppressor genes, and cancer drivers visualization. (a) The diagram shows representative oncogenes and tumor suppressor genes that accomplished Vogelstein’s criteria, according to the OncoKB database. Some genes are observed to play dual roles depending on the cellular context. Driver gene networks identified in head and neck cancer (b) and breast cancer (c) . The networks were generated using the DriverDBv4 tool ( http://driverdb.bioinfomics.org/ ) based on TCGA-US datasets for each cancer type. Each node represents a driver gene integrating information on somatic mutations, survival-associated expression, miRNA expression, DNA methylation alterations (hypo- or hypermethylation), and copy number variations (CNVs), obtained through bioinformatics tools. Yellow diamonds represent miRNAs, while red stars indicate data identified through multi-omics tools. Grey lines show the interactions between driver genes Key oncogenes, tumor suppressor genes, and cancer drivers visualization. (a) The diagram shows representative oncogenes and tumor suppressor genes that accomplished Vogelstein’s criteria, according to the OncoKB database. Some genes are observed to play dual roles depending on the cellular context. Driver gene networks identified in head and neck cancer (b) and breast cancer (c) . The networks were generated using the DriverDBv4 tool ( http://driverdb.bioinfomics.org/ ) based on TCGA-US datasets for each cancer type. Each node represents a driver gene integrating information on somatic mutations, survival-associated expression, miRNA expression, DNA methylation alterations (hypo- or hypermethylation), and copy number variations (CNVs), obtained through bioinformatics tools. Yellow diamonds represent miRNAs, while red stars indicate data identified through multi-omics tools. Grey lines show the interactions between driver genes Table 1 Representative oncogenes, tumor suppressor genes, and stability genes Name Cancer gene type Function Type of cancer* ABL Oncogene Tyrosine kinase Chronic Myeloid Leukemia FMS Oncogene M-CSF-receptor (tyrosine kinase) Sarcomas FOS; JUN Oncogene AP-1 regulator Osteosarcoma KIT Oncogene Stem Cell Factor-receptor (tyrosine kinase) Gastrointestinal Stromal tumors L-MYC Oncogene Transcription factor Lung N-MYC Oncogene Transcription factor Neuroblastoma H-RAS Oncogene GTP-binding protein Kidney K-RAS Oncogene GTP-binding protein Pancreatic; Colon N-RAS Oncogene GTP-binding protein Melanoma; Leukemias SRC Oncogene Kinase Colon; Sarcomas RB Tumor Suppressor Gene Transcription factor Retinoblastoma TP53 Tumor Suppressor Gene Transcription factor Half of all malignancies VHL Tumor Suppressor Gene Ubiquitination of HIFs Kidney APC Tumor Suppressor Gene β-catenin-binding Colon PTEN Tumor Suppressor Gene Phosphatase Glioblastoma; Lung WT1 Tumor suppressor Gene Transcription factor Wilm´s tumor STK11 Tumor suppressor Gene Kinase protein Ovarian; Pancreatic MUTYH Stability Gene Base excision repair Colon ATM Stability Gene DNA repair; cell division control Leukemias; Lymphomas; Brain BRCA1/2 Stability Gene DNA repair; cell division control Breast; Ovarian FANCA Stability Gene DNA repair Leukemias NBS1 Stability Gene DNA repair Lymphomas; Brain *Most cancer genes are involved in more than one type of cancer. In this table, only the most representative type(s) of cancer are shown Representative oncogenes, tumor suppressor genes, and stability genes *Most cancer genes are involved in more than one type of cancer. In this table, only the most representative type(s) of cancer are shown In most cancer types, including colon, breast, pancreas or brain, between 30 and 70 mutations are present per tumor [ 14 ]. Most of such mutations (> 90%) are single-base substitutions, whereas the rest are insertions or deletions of one or a few bases. Among the base substitutions, around 90% result in missense changes, 8% result in nonsense changes, and 2% in alterations of splice sites or untranslated regions immediately adjacent to the start and stop codons [ 14 ]. Interestingly, cancers, such as melanoma and lung, display around 200 mutations per tumor, attributed to mutagenic agents like UV light and cigarette smoke, respectively. Indeed, lung tumors from smokers show 10 times as many somatic mutations as those from non-smokers. Tumor mutational burden (TMB), mismatch repair deficiency (dMMR), and microsatellite instability (MSI) differ across various tumor types. This reaffirms that each tumor develops through very specific patterns. Kang et al. (2022) identified three types of tumors: (1) Tumors with high prevalence of dMMR and MSI and high TMB, such as endometrial and colorectal cancer; (2) Tumors with low prevalence of dMMR/MSI and high TMB, such as cervical cancer, head and neck cancer, and urothelial cancer; and (3) Tumors with low or moderate prevalence of all three markers, found in prostate and breast tumors, as well as sarcomas [ 19 ]. Without any doubt, somatic mutations are one of the main driving forces in cancer development. Two types of mutations can be identified based on their impact in tumor growth. Passenger mutations that do not affect tumor progression, and driver mutations that are critical for tumor progression by conferring a selective advantage to the transformed cell, including those resulting in single nucleotide variation (SNV) or copy number variation (CNA) [ 20 ]. These mutations contribute to refining the tumor’s abilities, such as immune system evasion, induction of angiogenesis, among other cancer characteristics. The mutational landscape is extremely diverse and is relevant in gene expression and in the differential drug sensitivity displayed by cancer cells, which, in turn, has relevant therapeutic implications [ 21 ]. It is no surprise that the trend in recent years is to treat diseases according to their molecular alterations rather than by the anatomical site of origin or histopathological features. The most robust somatic mutation databases, such as COSMIC, account for somatic mutations’ prevalence, diversity, three-dimensional structure, drug resistance, and actionable potential. Version v101 contains more than 24 million genomic variants, of which more than 5 million are variants in coding regions, more than 1 million are CNVs, and 17 million are variants in non-coding regions. The analysis of this mutational catalog has allowed the identification of 155 mutational signatures, the most frequent being single base substitutions (SBSs), with 95 signatures, as well as the identification of genes and mutations of relevance for cancer development, with 738 genes of which 79% are classified as Tier 1 ( https://cancer.sanger.ac.uk/cosmic ). This information is also relevant to the evolutionary construction of cancer from a mutational and chromosomal instability perspective. Fortunately, this information will increase thanks to deep sequencing technologies of multiple tumor areas, which would allow us to reconstruct the mutational evolution of the tumor more reliably. It has been estimated that somatic mutations have a frequency of 10 − 7 mutations/gene/generation or 10 –10 mutations/base pairs/cell generation and that, therefore, cancer-causing mutations are rare events [ 22 ]. Interestingly, the presence of driver mutations in non-malignant or normal conditions is relatively common, for example, in endometriosis and cerebral arteriovenous malformations, or simply due to advanced age. Some “RASopathies” (genetic conditions affecting the RAS-RAF-MAPK pathway) increase the risk of neoplasms such as myelocytic leukemia or neuroblastomas but do not increase the risk of melanoma, even though 75% of patients have mutated BRAF [ 23 ]. This could be related to mutational penetrance, defined as the extent to which a specific mutation manifests in the population exhibiting the associated phenotype. This suggest that mutational consequences depend on the spatiotemporal context in which they develop and raises the need to identify the conditions necessary for cancer development beyond driver mutations. Converging evidence indicates that the genomic evolution of cancer arises from a hybrid process in which stochastic variation coexists with constraining regulatory influences. In this framework, our use of the term “controlled” refers to the multiscale constraints that channel tumor evolution, rather than to a rigid or fully pre-programmed developmental sequence. Within the stochastic component, early mutations typically occur in genes involved in DNA replication, repair, or cell-cycle regulation, increasing genomic instability and enabling the accumulation of additional mutations and chromosomal rearrangements [ 24 ]. These alterations do not proceed in a deterministic, gene-to-gene cascade; instead, they reflect a heterogeneous accrual of somatic changes that subsequently undergo clonal selection and drift [ 25 ]. This evolutionary filtering provides a mechanistic explanation for why tumors with superficially similar genetic profiles may diverge in biological behavior and clinical outcome: only those subclones that acquire functionally advantageous combinations of alterations expand preferentially and ultimately dominate the tumor phenotype [ 26 ]. Multiple studies demonstrate that malignant transformation does not arise from a chaotic accumulation of mutations, but rather from hierarchically structured genomic events operating under strong selective pressure. High-throughput and single-cell sequencing analyses reveal that tumors are organized into phylogenetic trees with an early trunk of founding mutations, from which subclonal branches emerge in an order that is conditioned by the temporal acquisition of the initial driver events [ 27 ]. More recently, probabilistic frameworks such as Probabilistic Approach to Estimate the Temporal Order of Pathway Mutations Accounting for Intra Tumor Heterogeneity (PATOPAI) have enabled reconstruction of pathway-level mutation timelines by integrating intratumor heterogeneity and functional mutation annotation [ 28 ]. Complementarily, algorithms such as Identifying Mutation Order Pairs in Cancer (IMOP-Cancer) identify gene pairs in which the order of mutational events systematically reshapes tumor phenotypes, supporting the existence of preferential rather than random evolutionary routes [ 29 ]. Large scale chromosomal alterations driving tumor genome reorganization often follow reproducible patterns that are subject to selective pressures. Several tumors have been shown to accumulate somatic copy number alterations (SCNAs) and frequently undergo a whole genome duplication (WGD) event to tolerate chromosomal instability, thereby promoting the clonal expansion of specific SCNA configurations and structural rearrangements [ 30 ]. An example of selection driven genomic adaptation is the ability of the cellular microenvironment to shape the cancer karyotype. Tumor karyotypes are not fixed but rather plastic, and when cancer cells transition between environments such as from the patient to 2D culture, 3D organoid systems, or xenograft models the selective pressures shift, thereby determining which chromosomes are preferentially gained or lost [ 31 ]. These observations demonstrate that chromosome level alterations also evolve in a non-random manner, following selective rules imposed by the microenvironment. On the other hand, chromothripsis, defined by massive chromosomal rearrangements followed by “erroneous” reassembly, can generate recurrent selectable combinations of copy number states and gene fusions, reinforcing the notion that even the most dramatic genomic changes may arise as punctuated events within coherent evolutionary trajectories [ 32 ]. Watkins et al. described the evolution of SCNAs in various types of cancer. They found that loss of TP53 heterozygosity frequently precedes WGD, while CCNE1 gains follow WGD. They identified that 99% of tumors exhibit at least one subclona SCNAs. This is based on the identification of SCNAs that occurred before subclonal diversification of each tumor type, identifying the most recent common ancestor based on SCNA status. Alterations in SCNAs occur in an orderly, even parallel manner in various types of cancer. Genes such as BCL9, HIF1B, TERT, and MYC undergo alterations in separate clones in at least 37% of tumors. According to the authors, recurrent losses occur before WGD. Although the order of some SCNA events is described, the timing of these events is different for different tumor types. Thus, kidney renal clear cell carcinoma and glioma present SCNAs in the early stages, unlike other types of cancer [ 33 ]. The subclonal landscape results from positive selection, negative selection, or neutral evolution. Even extrachromosomal DNA can contribute to subclonal amplification events, contributing to genetic heterogeneity. Meanwhile, clonal SCNAs are more frequent in metastases than in primary tumors, where they are found as minor subclones. Consistent with increased proliferation, in tumors with chromosomal instability, clonal and subclonal SCNA burden correlates with increased cell cycle expression signature and mitotic index. They also found that particular chromosomal changes in different subclones within a single tumor are common in different cancer types, while clones with divergent mutations are less likely to survive. That is to say, similar chromosomal changes seem to evolve in parallel, suggesting the existence of specific pathways fundamental for tumor establishment and progression [ 33 ]. The accumulation of specific mutations demonstrates how cancer adapts and regulates its progression, following patterns that allow its survival and maximize its expansion. In recent years, new hypotheses on mutations and their impact in tumor progression have been presented. Heng proposed that cancer macroevolution is mediated primarily by karyotype changes rather than mutations, and suggests that alterations in karyotypes are, in fact, the main driving force in several types of cancer. His model implies a two-phase cancer evolutionary mechanism: macroevolution, mediated by karyotype or genome alterations, and microevolution, mediated by epigenetic alterations and gene mutations. According to Heng, elevated stress induces the so-called genomic chaos, defined as a rapid process of massive genome reorganization, that allows new survival systems for macroevolution; while microevolution facilitates modifications in the system that enable proliferation and competition [ 34 ]. It is possible that, during cancer development, many mutations are deleterious to the cells and, consequently, they are selectively eliminated. Traditionally, the study of mutations in cancer has focused on SNVs and CNVs. However, other approaches, such as that of Chen and coworkers, allow the identification of poorly characterized mutational patterns involving multiple base pairs and their functional impact. They identified 207 mutational patterns or rules in 27 types of cancer and their contribution to tumorigenesis. These rules are derived from the combination of codon alterations specific to each tumor type, based on a score generated through machine learning. Some of these rules are characteristic of specific tumor types, whereas some obey a single rule, such as rhabdoid tumors or clear cell renal carcinoma. In contrast, others have more than one rule, such as neuroblastoma with 14, and lung adenocarcinoma with 13. For example, a characteristic pattern in melanoma is CCC to CTC, TCC to TTC, and CCG to CTG, which involves the C > T change related to UV exposure present in more than 23% of patients. In esophageal cancer, the characteristic change is GCA to GAA, CCT to CAT, and TCG to TAG, i.e., C > A, which is present in more than 54% of patients. Another example is the TCG to TTG change, which involves the transition from Ser to Val, which are hallmarks of specific tumor types, such as in the RET gene in thyroid cancer. These changes reflect that specific changes are more relevant in certain types of cancer. Such mutational patterns were also functionally linked to biological behaviors such as tumor initiation or progression [ 35 ]. Although mutations are the genetic basis of cancer, and most of the hallmarks of cancer, proposed by Hanahan and Weinberg [ 6 ], can be explained in terms of various specific mutations, non-mutational mechanisms, such as epigenetic mechanisms, are also involved and play important roles in tumorigenesis.

Since the work by Rudolph Virchow, in the 19th century, cancer has been regarded as uncontrolled cellular growth that can lead to the formation of aberrant tissue masses, known as tumors, and, eventually, to death of the individual. However, evidence reported over the past 30 years, based on genomic, epigenetic, molecular, and cellular studies, indicate that, although some stochastic molecular events take place at different steps during the tumorigenic process, cancer is, for the most part, a firmly controlled process. There seems to be no doubt that cancer cells follow specific biological rules that have been called the hallmarks of cancer [ 6 ]. Several findings support the notion that cancer is not an uncontrolled process. As discussed herein, (i) specific patterns can be observed within the vast array of genomic, epigenomic, and molecular alterations that drive malignant transformation of normal cells into cancer cells. (ii) There is a correlation between the type of molecular alterations found in transformed cells and the type of cancer developed; indeed, carcinogenic alterations are not the same in all types of cancers. (iii) It is clear that not all cancer cells within a tumor are similar; there is heterogeneity (not all cancer cells are able to generate a tumor, and not all cancer cells are able to metastasize). In fact, there seems to be a hierarchy within the neoplastic cells. (iv) The process by which cancer cells spread to distant tissues (including the EMT) is a finely regulated process. Finally, (v) metastasis is not random, but there seems to be specific patterns of directionality. An intriguing aspect of cancer biology is the ability of tumor cells to replicate and survive for periods of time that usually exceed those of normal cells (“replicative immortality”) [ 6 ]. Such a cell immortality is defined as an excessive number of cell divisions that exceed the Hayflick limit, which is about 40–60 generations in vitro. This feature of cancer cells is influenced by increased telomerase activity [ 196 ]; however, several other elements, such as reduced apoptosis and autocrine control of cell division, are also involved in such an “immortality”. Based on physics and biology principles, it has been established that highly unstable (chaotic) systems are unlikely to survive indefinitely. In contrast, it has been suggested that increase in the complexity and connectivity of biological networks can enhance their resistance to endogenous/exogenous perturbations, and chaos [ 197 ]. On the other hand, the karyotypic theory of immortality argues that carcinogenesis is a form of speciation, in which chromosomal rearrangements and aneuploidies, initially random, generate karyotypic diversity that undergoes selection and stabilizes toward an independent, autonomous replicative karyotype, even in the absence of telomerase activity [ 198 ]. On the other hand, the karyotypic theory of immortality argues that carcinogenesis is a form of speciation, in which chromosomal rearrangements and aneuploidies, initially random, generate karyotypic diversity that undergoes selection and stabilizes toward an independent, autonomous replicative karyotype, even in the absence of telomerase activity [ 198 ]. This argument supports the notion that tumors are biological systems formed by complex molecular and cellular interconnections that work under the action of multiple regulatory mechanisms, so that they can be sustained over multiple replicative rounds. Hence, tumors are far from being chaotic biological systems running out of control. Understanding the genomic, molecular, and cellular rules under which cancer cells originate and develop is crucial not only to gain a deeper insight into the biology of this group of diseases, but to develop diagnostic, prognostic, and therapeutic approaches and strategies. For instance, it is now clear that CSCs play a key role in carcinogenesis and relapse, and that such cells differ from the vast majority of the cells that form the tumor mass, which are usually vulnerable to chemotherapy and radiation. Thus, developing tools and tactics for identifying and eliminating CSCs in a selective way will result in more accurate therapies [ 118 ]. The vast majority of cancer-associated deaths are the result of metastatic activity. Thus, dissecting and deciphering the elements and mechanisms leading to tumor cell invasion and metastasis will lead us to the development of more effective anti-metastasis treatments [ 195 ]. In summary, cancer is the result of an intricate network of abnormally expressed genes, abnormally produced proteins, and abnormally delivered molecular signals that lead to the development of transformed cells that abnormally expand, interact with their surrounding microenvironment, and disseminate. Tumor development involves an extremely wide conglomerate of genomic, molecular, and cellular elements that interact, in an orchestrated manner, to favor the growth of the transformed cells and their progeny at the expense of normal cells (Fig.  5 ). Cancer implies abnormal cellular growth, no doubt about it; but not uncontrolled growth. In fact, considering cancer as uncontrolled cellular growth would be underestimating and misunderstanding such a complex, assembled malignant process. Fig. 5 Cancer is a controlled biological process. Rather than being a chaotic phenomenon, cancer represents a complex and meticulously orchestrated biological process. This illustration employs the analogy of a Rubik’s Cube to exemplify how several pieces such as cellular proliferation, cancer stem cells (CSCs), clonal events, spatial reconfiguration of the genome, chromatin compartmentalization, oncohistone profiles, mutations, epithelial-to-mesenchymal transition (EMT), regulation by ncRNA (non-coding RNA). angiogenesis, and metabolic changes, among others— fit together precisely to transform a healthy cell into a malignant one. Just as a Rubik’s Cube does not assemble by chance, cancer is not a random occurrence -although some stochastic events take place along the tumorigenic process- but the result of a series of perfectly coordinated events. The yellow cubes represent processes that remain to be elucidated, adding to the complexity of cancer Cancer is a controlled biological process. Rather than being a chaotic phenomenon, cancer represents a complex and meticulously orchestrated biological process. This illustration employs the analogy of a Rubik’s Cube to exemplify how several pieces such as cellular proliferation, cancer stem cells (CSCs), clonal events, spatial reconfiguration of the genome, chromatin compartmentalization, oncohistone profiles, mutations, epithelial-to-mesenchymal transition (EMT), regulation by ncRNA (non-coding RNA). angiogenesis, and metabolic changes, among others— fit together precisely to transform a healthy cell into a malignant one. Just as a Rubik’s Cube does not assemble by chance, cancer is not a random occurrence -although some stochastic events take place along the tumorigenic process- but the result of a series of perfectly coordinated events. The yellow cubes represent processes that remain to be elucidated, adding to the complexity of cancer

From the epigenome approach, three types of genes can be identified in cancer development: modulators (genes that activate or repress the epigenetic machinery), modifiers (genes that modify DNA methylation or chromatin structure and that can be mutated or not), and mediator genes (genes that are regulated by epigenetic modifiers and that increase pluripotency or survival) [ 36 ]. Non-coding RNAs, on the other hand, have proven to be fundamental in the development and progression of cancer, with some also playing an important role in epigenetic regulation. For the last 25 years, the existence in human cells of a wide and heterogeneous group of RNA transcripts that are not translated into proteins has been clearly demonstrated. Such a family of molecules, known as non-coding RNAs (ncRNAs), are important regulators of multiple biological functions, and alterations in their structure/expression/function have been implicated in several diseases [ 37 ]. Based on their length, two subfamilies of ncRNAs have been identified: small ( 200 nucleotides) RNAs. Within the former, there is a group, known as micro RNAs (20–23 nucleotides; miRNAs), whose members have been identified as key regulators of gene expression (Fig. 2 i). They bind, by specific base-pairing, to a variety of target messenger RNAs (mRNAs), and in this way, they repress their translation. Three mechanisms have been identified for such a repression: cleaving the mRNA strand into two fragments; destabilizing the mRNA by shortening its poly A tail; or reducing the actual translation of the mRNA into protein [ 38 ]. Some examples of miRNAs are shown in Table 2 . The human genome encodes around 2000 miRNAs, and many of them have been implicated in the ethology of cancer [ 39 ]. In this regard, it is important to point out that there are specific patterns of altered expression of miRNAs in certain types of cancer. For instance, increased expression levels of miR-205 and miR-373 have been found in colorectal cancer [ 40 ]; increased cell proliferation in hepatocellular carcinoma seems to result from the interaction of miR-21 with the tumor suppressor gene MAP2K3 [ 41 ]; and downregulation of miR-506 seems to be involved in the development of cervical cancer [ 42 ]. Interestingly, it has been found that cell-free circulating miRNAs are stable in blood, and their levels are increased in most types of cancer; thus, they have been used as potential biomarkers [ 43 , 44 ]. miRNAs have also been shown to significantly impact invasion and metastasis processes [ 45 ] Within these mechanisms, some miRNAs act as oncogenes, while others function as tumor suppressor genes. For instance, in the growth of primary tumors, miR-17/92, miR-222, and miR-211 play roles as tumor suppressors, whereas miR-200 and let-7 function as oncogenes. In invasion processes, miR-9, miR-10b, and miR-107 exhibit tumor-suppressive functions, while miR-29c, miR-34a, miR-214, miR-340, among others, act as oncogenes. Interestingly, they can exhibit opposite roles depending on the type of cancer. However, due to the diversity of potential target genes and the timing of their effects, it is challenging to determine their precise relevance in the sequence of events leading to cancer development [ 45 ]. The long non-coding RNAs (lncRNAs), on the other hand, seem to have more complex mechanisms of action, including the formation of networks of ribonucleoprotein (RNP) complexes. Like miRNAs, they can act as oncogenes or tumor suppressors. They play a crucial role in maintaining pluripotency and participate in several functions related to cancer development and progression, significantly influencing gene expression, cellular signaling pathways, and the dynamics of the tumor microenvironment [ 46 ]. Interestingly, they can act as molecular enhancers, decoys, scaffolds, and guides. Indeed, lncRNAs can function as modular scaffolds to specify higher order organization in RNP complexes and in chromatin states [ 46 , 47 ]. Some lncRNAs can interact with chromatin remodeling complexes, which regulate the accessibility of DNA for transcription. By binding to these proteins, lncRNAs have the ability to alter epigenetic marks, such as DNA methylation or histone modifications, changing the expression of tumor suppressor genes or oncogenes (Fig. 2 ii). These mechanisms occur at specific times and locations within the cell. lncRNAs are so crucial that they have been considered “master regulators” in the initiation, maintenance, and regulation of CSCs derived from different neoplasms. Examples of lncRNAs that have stood out for their role in CSCs include CUDR, Lnc34a, Linc00617, CTCF7, ROR, DILC, HOTAIR, H19, HOTTIP, ATB, HIF2PUT, PVT1, SOX2OT, MALAT-1, DYNC2H1-4, SOX4, and ARSR Uc.283-plus, which promote the tumors adaptation to the changing conditions of the tumor microenvironment [ 48 ]. A significant number of lncRNAs show altered expression in several types of cancer; thus, they have been implicated in tumorigenesis [ 8 , 49 ]. For instance, ANRIL has been linked to prostate cancer; XIST to different female tumors; HOTAIR to breast cancer; and KCNQ1OT1 to colorectal cancer [ 50 ]. Additionally, it has been observed that certain lncRNAs are activated in response to viral agents, such as HPV, directly or indirectly contributing to the formation and progression of cancer [ 9 ]. Another example is in triple‑negative breast cancer (TNBC), lncRNA-SOX9-AS has been shown to exhibit high expression levels, while the expression of Lnc-PXDN-3:1, Lnc-SYDE, and LINC01087 is reduced. These lncRNAs could contribute to the tumor aggressiveness of this cancer subtype. Furthermore, low expression LINC01087, LINC02568, ACO22196 , and Lnc-EGOT is associated with reduced overall patient survival [ 51 ]. Additionally, unique lncRNAs specific to Mexican patients with TNBC, such as LINC00174, have been identified and may influence survival outcomes [ 51 ]. This reaffirms that the variability in lncRNA expression among different tumors highlights the complexity and unique nature of each cancer type. Another example of the highly specific impact of lncRNA on breast cancer tumors is its role in maintaining genomic instability. Bao and colleagues identified that LINC02207 and RP11-358L4.1 are capable of stratifying patients into high- and low-risk groups, with significantly different survival outcomes, including those with wild-type TP53. This was also validated in ovarian cancer, where a correlation with the genomic mutation rate was observed [ 52 ]. Other representative examples of LncRNAs are presented in Table 2 . Fig. 2 Interaction between nuclear organization and epigenetic mechanisms in cancer control. The figure depicts the nucleus of a cell, highlighting chromosomal territories and four magnified views that illustrate distinct layers of epigenetic regulation as well as the participation of ncRNAs. ( i ) In the cytoplasm, the action of a miRNA that modulates gene expression through mRNA degradation is illustrated. ( ii ) Mechanism of action of a lncRNA that recruits chromatin-modifying complexes to regulate the expression of neighboring genes. ( iii ) Oncohistones mutations affecting key sites for post-translational modifications, disrupting interactions with chromatin remodeling complexes and promoting cancer-specific epigenetic programs. ( iv ) Chromatin compartmentalization and the formation of TADs mediated by CTCF; ( vi-a ) transition of TAD architecture from a healthy to a cancer cell Interaction between nuclear organization and epigenetic mechanisms in cancer control. The figure depicts the nucleus of a cell, highlighting chromosomal territories and four magnified views that illustrate distinct layers of epigenetic regulation as well as the participation of ncRNAs. ( i ) In the cytoplasm, the action of a miRNA that modulates gene expression through mRNA degradation is illustrated. ( ii ) Mechanism of action of a lncRNA that recruits chromatin-modifying complexes to regulate the expression of neighboring genes. ( iii ) Oncohistones mutations affecting key sites for post-translational modifications, disrupting interactions with chromatin remodeling complexes and promoting cancer-specific epigenetic programs. ( iv ) Chromatin compartmentalization and the formation of TADs mediated by CTCF; ( vi-a ) transition of TAD architecture from a healthy to a cancer cell Table 2 Representative noncoding RNAs implicated in carcinogenesis Name Type of ncRNA Alteration/type of cancer let-7a-5p miRNA Downregulated in breast and gastric cancer miR-223-3p miRNA Overexpressed in gastric and prostate cancer miR-134 miRNA Downregulated in colon and lung cancer miR-141 miRNA Overexpressed in prostate and ovarium cancer miR-146a miRNA Downregulated in lung and colon cancer mi-R-221 miRNA Overexpressed in colon and lung cancer mi-R-222 miRNA Overexpressed in colon and lung cancer mi-R-223 miRNA Overexpressed in liver and lung cancer mi-R-210 miRNA Overexpressed in B cell lymphoma, lung, and pancreatic cancer ANRIL lncRNA Upregulated in prostate cancer XIST lncRNA Downregulated in breast, ovarian, and cervical cancer HOTAIR lncRNA Overexpressed in breast cancer KCNQ1OT1 lncRNA Loss of imprinting in colorectal cancer H19 lncRNA Upregulated in gastric cancer SRA lncRNA Upregulated in breast cancer TERRA lncRNA Downregulated in several cancer cell lines MALAT1 lncRNA Promotes migration and metastasis in lung cancer PTENP1 lncRNA Lost in many human cancers *Most ncRNAs may be involved in more than one type of cancer. In this table, only the most representative type(s) of cancer are shown Representative noncoding RNAs implicated in carcinogenesis *Most ncRNAs may be involved in more than one type of cancer. In this table, only the most representative type(s) of cancer are shown The fundamental unit of chromatin is the nucleosome, which is composed of a histone octamer wrapped around 147 base pairs of DNA [ 53 ]. It is well known that transcriptional regulation is closely related to a profile of chemical modifications in particular amino acid residues of histones, known as histone codes. Interestingly, tumors have now been identified as exhibiting specific oncohistone profiles; that is, mutations in particular amino acid residues that are consistent and recurrent under certain conditions, such as specific cancer types. These mutations can alter gene expression and contribute to cancer development. It has been identified that oncohistones mutations are not random. Instead, these mutations occur in globular domains, which are important for nucleosome integrity and stability. Furthermore, they are primarily located in the N-terminal tail of histone H3, close to residues that are susceptible to relevant post-translational modifications [ 54 , 55 ]. These alterations also affect the interaction of chromatin remodeling complexes, often disrupting their ability to properly modify chromatin (Fig. 2 ). Numerous studies have established strong correlations between specific histone mutations and distinct cancer types. For example, K27M mutation in histone H3 is characteristic of diffuse intrinsic pontine gliomas, and the G34V/R mutation in H3 has been observed in pediatric glioblastomas. Similarly, histone H3.3 mutations, such as K27M and G34R/V, are hallmarks of childhood brain tumors [ 56 ]. In chondroblastomas, the histone variant H3.3, encoded by the H3F3B gene, is mutated at K36M, and in giant cell tumors of bone, at G34W/L. Interestingly, mutations have been found in the genes encoding histones H2A and H2B in uterine and ovarian carcinosarcomas (Fig. 2 -iii). Experiments have shown that this mutation induce the expression of EMT markers, in addition to enhancing tumor migration and invasion processes [ 57 ]. Beyond these examples, histone mutations may also play significant roles in other cancers. For instance, mutations in histone H1 are frequent in diffuse large B-cell lymphomas. Likewise, in squamous cell carcinomas of the head and neck, a K36M mutation in histone H3 has been reported, preventing the activity of Nuclear receptor-binding SET Domain protein 1 (NSD1), which is responsible for the mono- and di-methylation of lysine 36 of H3. This results in the loss of proper histone H3 methylation at K36, blocking cell differentiation, inducing high levels of Long Interspersed Nuclear Element-1 (LINE1)-mediated retrotransposition, which can lead to genome instability, and ultimately promoting oncogenesis [ 55 , 58 , 59 ]. While oncohistone profiles have traditionally been determined in tumor tissue samples, recent studies have demonstrated the potential for assessing the unique epigenetic profile of circulating cell-free nucleosomes (cfNs) in plasma. These cfNs, which reflect the epigenetic status of tumor cells, enable the differentiation of patients with various cancer types, such as colorectal cancer, pancreatic ductal adenocarcinoma, and diffuse midline glioma (DMG), as well as healthy subjects [ 60 ]. For example, in the case of DMG, the H3-K27M mutation in histone H3 has been identified in patient plasma. This is relevant because the current classification of DMG is based on the status of H3-K27 and the H3-K27M mutation. This criterion has a direct impact on patient prognosis, with patients with the H3-K27M variant having a worse outcome compared to those with the wild-type variant (H3-K27) [ 61 ]. Recurrent mutations in oncohistone ‘hotspots’ (regions of frequent mutation) demonstrate that the epigenetic machinery plays a defined and key role in altering specific processes related to tumor oncogenesis; thus, refuting the idea of cancer as a purely random disorder. According to various studies, the biochemical properties of oncohistones translate into pathological alterations within the epigenetic landscape. For example, these mutations disrupt the balance of histone acetylation and methylation levels, thereby impacting chromatin remodeling. Progress in techniques such as computational modeling, biochemistry, high-resolution microscopy, genome editing, and chromosome mapping has characterized the spatial organization of chromatin into distinct compartments or territories. These structures, encompassing permissive and repressive regions, self-associating domains, and regulatory loops, are fundamental to maintaining cellular integrity and regulating transcription (Fig.  2 ). The three-dimensional (3D) architecture of chromatin has been shown to influence several hallmarks of cancer, such as genomic instability, the maintenance of proliferative signaling, invasion, metastasis, angiogenesis, and inflammation, among others [ 62 ]. To understand how this 3D organization impacts these tumor processes, it is essential to explore how the genome is spatially structured within the nucleus. High-throughput chromatin conformation capture techniques (Hi-C) at 1 Mb resolution have revealed that the human genome is organized into spatially segregated chromosomal territories, which are further divided into active (A) and inactive (B) compartments within the cell nucleus. At the sub-Mb scale, chromatin is organized into topologically associated domains (TADs), characterized by preferential intra-domain interactions. TADs often contain smaller contact subdomains, called chromatin loops, that generally correlate with gene activation [ 63 ]. The formation of TADs is not a random process, but rather one that is carefully orchestrated and written into the epigenome. TADs are primarily structurally formed by proteins such as CCCTC-binding factor (CTCF), a highly conserved zinc finger protein that binds to specific DNA sequences located at the TAD borders. CTCF binds to DNA and, upon dimerization, acts as an “anchor” connecting distant genomic regions, thereby limiting interactions between elements of adjacent TADs (Fig. 2 iv-iva). This organization is precisely encoded, ensuring the functional segregation of the genome into distinct spatial regions. CTCF-mediated interactions are also key to the regulation of gene expression, maintaining chromatin stability and order. In addition to CTCF, other proteins are important in TAD formation, such as cohesin and Yin Yang factor 1 [ 64 ]. When a healthy cell transforms into a cancerous cell, profound changes occur in the 3D organization of the genome. This restructuring affects the way DNA folds and organizes in the nucleus, leading to new and aberrant interactions between genomic regions that normally do not interact. As a result of these structural changes, a malignant transformation occurs in which oncogenes are activated and tumor suppressor genes are silenced due to new interactions between enhancers and promoters, contributing to changes in gene expression that drive cancer development and progression. Studies have shown that chromosomal territories are dynamic and can undergo spatial repositioning within the nucleus of a cancer cell. For example, significant modifications in chromosome position have been described in myeloma patients. In situ hybridization assays have revealed that chromosomes 4, 9, 14, and 18 tend to shift toward the center of the nucleus in myeloma cells. In contrast, chromosomes 11 and 16 are located more peripherally compared to normal lymphocytes [ 65 ]. Consequently, these altered spatial arrangements could contribute to the aberrant gene expression and function of myeloma cells. Another example of this type of chromosomal rearrangement occurs in breast cancer cells (MCF7), where chromosomes 16 to 22 (chr16-22), characterized by being small and rich in genes, show a lower frequency of interchromosomal interactions with each other, compared to healthy epithelial cells (MCF-10 A) [ 66 ]. Notably, a significant 12% of the genomic compartments in MCF-10 A changed from type A to B and vice versa in MCF-7 cells. Furthermore, the authors observed a transition from compartment B to A in cancer cells on chr16–22, indicating a higher frequency of open compartments in MCF-7 chromosomes. These findings suggest the existence of an intricate interplay between interchromosomal organization, compartmentalization, and gene expression, which allows specific genes to be activated according to the tumor context. In MCF-7, pathway enrichment analysis of genes located in altered compartmental regions (chr16-22) reveals repression of WNT signaling [ 66 ]. Hi-C experiments have further revealed that chromosome 1 (Chr1) in oral squamous cell carcinoma (OSCC) cells exhibit an increase in short-range interactions and a reduction in long-range interactions compared to healthy cells. Concurrent alterations in chromatin compartmentalization (A and B) are also observed in OSCC cells, affecting 28 genes functionally related to the cell cycle, proliferation, and cell adhesion, including MYO1B, STC1, CDKN2A, HMGA2, RBBP8, and PRNP [ 67 ]. Emerging evidence suggests that transcription factors play a significant role in the B-to-A compartment switch. In a squamous cell lung cancer cell line, the transcription factor KLF5 was shown to interact with the remodelers SWI/SNF, TOP2B, and CBP. Knockout of KLF5 reduces the recruitment of these complexes to squamous cell lung cancer-associated genes (ID1, TP63, FOXE1, PDGFA, WNT10A, and KRT5), suggesting that KLF5 is involved in DNA compartmentalization and in changes in gene expression and cellular properties [ 68 ]. Regarding TADs, it has been observed that they exhibit different profiles depending on the type of cancer. For example, in breast and prostate cancer, the acquisition of new TAD boundaries has been reported, which is usually accompanied by an increase in the total number of TADs and a reduction in their size (Fig. 2 i-ia) [ 66 , 69 ]. Strikingly, a large proportion of newly established TAD boundaries in cancer tend to coincide with regions exhibiting copy number variation. A notable example is found in prostate cancer, where a recurrent deletion (14.8%) at 17p13.1, encompassing the TP53 tumor suppressor gene locus, leads to the fragmentation of a single TAD into two smaller TADs [ 69 ]. In contrast, in colorectal cancer and acute lymphoblastic leukemias, the weakening or even disappearance of TAD boundaries is more common, indicating distinctive structural changes in these tumor types [ 70 , 71 ]. Within the so-called “molecular chaos of cancer”, our current understanding of the finely regulated organization of epigenetic elements has led to the development of drugs targeting the three-dimensional structure of chromatin. One example is curaxins (CBL0137), which intercalate into DNA without causing detectable damage. These molecules can unwind DNA from nucleosomes, induce partial histone eviction, and promote DNA transitions from the B to the Z form. As a result, they significantly alter the spatial organization of the genome, effects that are at least partly mediated by the depletion of CTCF at its binding sites [ 72 ]. The genetic and epigenetic alterations manifested in the transformed cell, as previously described, trigger a cascade of cellular modifications that culminate in the malignant neoplastic phenotype, affecting both its structure and functionality. This intricate interplay has given rise to one of the most paradigmatic questions in this field: what comes first in this sequence of events, genomic changes or epigenomic ones?

Today, cancer represents the second cause of death worldwide [ 1 , 2 ]. According to the International Agency for Research on Cancer (IARC), approximately 20 million new cases of cancer are reported annually, and 10 million people die from a particular type of cancer. It is projected that in the year 2030, 24 million new cases will be detected, and around 13 million deaths will occur [ 3 , 4 ]. IARC has further predicted that in the year 2040, 27 million new cases will be detected, and almost 14 million deaths will occur [ https://gco.iarc.fr/tomorrow/en ]. Thus, without any doubt, cancer is one of the main public health priorities worldwide [ 5 ]. Cancer has been defined as a group of diseases that result from the uncontrolled growth (i.e., proliferation and differentiation) of abnormal cells that invade nearby tissues and spread to other parts of the body, generating physiological alterations and organ failure that could lead to the death of the individual (National Cancer Institute Web page: https://www.cancer.gov/about-cancer/understanding/what-is-cancer ). Although this is a widely accepted definition, commonly used in the scientific and academic arenas, the notion of “uncontrolled” growth would imply that cancer cells grow erratically, following no biological rules and without any regulation. This, however, does not seem to be the case. Evidence presented over the past three decades, based on genomic, epigenetic, molecular, cellular, immunological, and physiological approaches, indicate that cancer is, for the most part, a controlled, yet abnormal, process that results in the formation of solid tumors or hematologic neoplasms. In other words, cancer cells do follow biological rules which are different from the rules followed by normal cells. Such “abnormal” rules have been regarded as the hallmarks of cancer [ 6 ]. Although it is clear that stochastic molecular events occur throughout the tumorigenic process, several findings support the concept that cancer is not an uncontrolled process. First, specific patterns can be observed within the vast array of genomic, epigenomic, and molecular alterations that drive malignant transformation of normal cells into cancer cells. Second, there is a correlation between the type of molecular alterations found in transformed cells and the type of cancer developed; indeed, such alterations are not the same in all types of cancers. Third, it is clear that not all cancer cells within a tumor are equal; there is heterogeneity (not all cancer cells are able to generate a tumor, and not all cancer cells are able to metastasize). In fact, there seems to be a hierarchy within the neoplastic cells. Fourth, the process by which cancer cells spread to distant tissues (including the epithelial-mesenchymal transition [EMT]) is a finely regulated process. Finally, metastasis is not random, but there seems to be specific patterns of directionality. Within this context, herein, we present an up-to-date comprehensive overview on the current evidence supporting the notion that, although cancer implies abnormal -malignant- cellular growth, it is not uncontrolled cellular growth. We summarize and discuss the evidence indicating that cancer originates and progresses under the action and influence of specific factors and mechanisms that regulate, in a finely way, tumor cell growth and dissemination.