Given this background, it would appear that compelling evidence exists that cancers arise from accumulated defects affecting multiple genes and pathways, and precancerous (benign) precursor lesions harbor fewer of the key gene defects. One question, therefore, is how many rate-limiting defects or “hits” are required for cancer development. Although a definitive answer to this question cannot be given at this point, some estimates can be offered. Most common cancers show dramatically increased incidence with increasing age. On the basis of analysis of the age-specific incidence of a number of common cancers and some straightforward assumptions about the rate of mutations and the size of the target cell population, it was argued as early as the mid 1950s that most common epithelial cancers arise as the result of four to seven rate-limiting events.1,2 It was inferred that these rate-limiting events might represent mutational events. Moreover, benign lesions were inferred to arise as the result of fewer gene defects, consistent with the fact that recognizable benign lesions that are often found show an age-incidence distribution that was shifted roughly one to two decades earlier in life than cancers arising in the corresponding organ or tissue sites. Nevertheless, confounding the use of age-incidence data to model the number of rate-limiting mutations were questions about certain key biological assumptions underlying the multihit models. Given the attendant uncertainties with estimates of rate-limiting mutation numbers derived largely or solely from age-incidence data and the practical difficulties in defining the nature and significance of all inherited and somatic gene defects in cancer, a definitive answer to the number of rate-determining defects in a particular cancer type has not yet been obtained. For instance, even one of the early comprehensive molecular analyses of breast and colon cancer indicated that dozens to hundreds of somatic mutations were often present in a given cancer,³ without even considering the even far greater number of epigenetic and gene expression changes present in the cancer cells relative to adjacent normal cells.

Study of the genetic alterations of pediatric cancers that occur much earlier and are perhaps less complex in terms of their genetic signature may offer insights into key cancer target genes (see Chapter 95 on pediatric solid tumors).

Clonal Selection and Evolution in Cancer

As previously noted, most if not all cancers arise from preexisting precancerous populations of cells, and multiple genetic and epigenetic defects are likely needed for conversion of a normal cell to a cancerous cell. Molecular studies of cancers of various types and their corresponding associated precancerous lesions have yielded some fundamental insights into the processes likely to be critical in the emergence of the cancer. First, although normal tissues and tissues from noncancerous disease states display polyclonal (balanced) cell populations, the neoplastic component that is present in benign lesions and cancers displays a clonally related cell population, consistent with the notion that neoplastic transformation of one or at most a few cells within a tissue give rise to all daughter cells that are present in the tumor. Second, in tumors in which it has been possible to analyze both cancer cells and associated precancerous cell populations, a subset of the somatic gene defects present in the cancer are clonally represented in the precancerous cell population. Other somatic gene defects appear to be acquired during progression from the precancer subclones to the dominant subclone in the cancer.

These molecular findings in benign and malignant tumors are essentially consistent with a model initially proposed by Foulds ⁴ and subsequently advanced by Nowell⁵ (Fig. 13-1). In brief, the clonal evolution model predicts that cancers arise as the result of successive expansions of clonally related cell populations. The successive expansions are driven by the gradual or punctuated acquisition of mutations and gene expression changes that endow a particular cell and its progeny with a selective advantage over cells that do not harbor the gene defects. In essence, clonal selection is an evolutionary process that allows the outgrowth of precancerous and cancerous cells that carry mutations and gene expression changes that confer the most potent proliferative and survival properties on the cancer cells. It is important to note that the specific constellation of genetic and epigenetic changes that are present in precancerous and cancerous cells is context-dependent and certainly varies considerably from one cancer type to another and even most certainly varies to a significant degree among patients whose cancers display quite similar clinical and histopathological features. The basis for the context-dependent relationship of the changes that confer a selective growth advantage in a particular cancer may reflect physiological differences in organ site and the tissue microenvironment within the organ site, the identities of the preceding somatic gene defects in the precancerous or cancerous clone, and even the constitutional sequence variations and particular gene expression patterns that are present in nonneoplastic tissues of a given patient. This issue of context-dependent effects of gene defects that promote clonal selection in neoplastic cells will be addressed further in the following sections.

Figure 13-1 Role of clonal selection in cancer development and progression. Clonal selection is essentially an evolutionary process that promotes outgrowth of precancerous and cancerous cells that carry the mutations and gene expression changes that confer the most potent proliferative and survival properties on the cancer cells in a given context. The schematic diagram indicates that the stepwise emergence of benign and malignant cells over time is critically influenced by mutations and epigenetic defects. Neoplasms most likely arise from a stem cell or progenitor cell population that is capable of additional cell divisions and acquisition of certain differentiated characteristics. After the accumulation of a particular constellation of mutations and epigenetic defects in oncogenes and tumor suppressor genes, a successful malignant subclone will outgrow the various competing neoplastic subclones. Further genetic heterogeneity within a malignant subclone (not depicted) is likely, and the genetic heterogeneity might give rise to new subclones that display increased invasive and metastatic potential. (Modified from Kern SE. Progressive genetic abnormalities in human neoplasia. In: Mendelsohn J, Howley PM, Israel MA, Liotta LA, editors. The molecular basis of cancer. 2nd ed. Philadelphia: Saunders; 2001.)

The clonal evolution model has biological and clinical ramifications, just a few of which will be mentioned here. First, the clonal gene defects that are present in a cancer can be traced in precancerous lesions from the same organ site with a goal of attempting to clarify the preferred order in which gene defects arise in the natural history of a particular cancer type. The particular order in which defects accumulate during the initiation and subsequent progression of one cancer type often differs from that in another cancer type. As a result, a genetic or epigenetic change that is critical in tumor initiation in one cancer type might contribute to tumor progression in another tumor type and vice versa. Second, defects that arise at “early” stages of tumorigenesis might play a vital role not only in tumor initiation but also in the aggressive behavior of advanced stage cancers. Third, the model predicts that the acquisition of further genetic heterogeneity will be a common and important factor in primary cancer lesions and metastases. Genetic heterogeneity likely plays a significant role in resistance to chemotherapy and the emergence of aggressive cell populations in patients with advanced cancer.

Recent comprehensive sequence-based analyses of cancer cell populations from individual patients in which the tumor cells under study were distinct from one another in location and/or time has begun to reveal that quite dramatic intratumoral genetic heterogeneity may be a “rule” in cancer, rather than an exception.⁶ Certain, perhaps key, initiating genetic lesions might be shared among all neoplastic clones, but geographically distinct regions of large primary tumors may have very distinct mutation profiles from those in other portions of the primary tumor, and the metastatic cell populations may have considerable mutational divergence from the nonmetastatic cells. As such, it seems that quite extensively branched evolutionary growth may be an important feature in both primary tumors and metastatic lesions, with multiple competing clonal populations evolving through divergent and convergent mutational mechanisms.⁶ This more recent view of the potentially quite extensive intratumoral genetic heterogeneity in any given cancer and the contributions of intratumoral heterogeneity to tumor progression contrasts with some earlier views. Prior to the recent studies, it was suspected that the cell populations in many primary cancers might be more homogeneous, where somatic mutations were accumulated in a more stepwise fashion as a result of multiple sequential clonal sweeps of each variant cell populations in the primary cancer, with metastases perhaps most often arising from the clonally dominant cell population in the primary tumor.

It is important to note that clonal somatic mutations are often presumed to have a causal role in promoting further tumor outgrowth or progression because somatic mutations can become clonal (i.e., present in all neoplastic cells) by only a limited number of mechanisms. For instance, the genetic alteration itself could have been selected for because it provided the neoplastic cell with a growth advantage, allowing it to become the predominant cell type in the tumor (clonal expansion). Genes with critical roles in promoting clonal outgrowth in a given cancer have been termed drivers.⁷ Alternatively, a somatic mutation, when detected, might have arisen essentially coincident with another, perhaps undetected, alteration that was the crucial change underlying clonal outgrowth. Somatic mutations of this latter type have been termed passengers.⁷ Genes that are mutant in a significant fraction of cancers and for which other lines of evidence link them to the cancer process can more readily be classified as drivers. However, on the basis of early data from some large-scale sequencing analyses that reveal large numbers of distinct genes that are each mutated in only a minority of cancers of a given type,^3,8–10 sorting out drivers and passengers might not be entirely straightforward, based solely on sequencing data, but will likely require a significant body of functional studies and data. Statistical arguments may fall short, at least before large volumes of sequencing are obtained for the alterations that are private or rare, yet still biologically important when present in a given patient’s cancer.

Given this discussion about the potential uncertainties associated with linking specific somatic mutations to cancer development, it is perhaps apparent why it is even more challenging to assign causal significance to any given gene expression change in precancerous and cancerous cells. In particular, the ambiguities in assigning a causal role to gene expression and epigenetic changes in the absence of mutational or functional data that might inform the situation are in large part due to the difficulties in determining whether a gene expression or epigenetic change merely reflects or is causally involved in the cancer process. Nonetheless, if a specific gene or epigenetic alteration can be shown to promote tumorigenesis or neoplastic transformation in a variety of in vitro tumor models and animal model experiments or if the same gene or chromosomal region is recurrently altered by particular epigenetic changes in tumors, then it might be reasonable to infer that the particular defect might indeed have a causal (driver) role in tumorigenesis.

Contribution of Gene Defects to the Signature Traits of Cancer Cells

Cancer represents a highly heterogeneous collection of diseases. Each cancer type has distinct biological and clinical features and a variable prognosis. Even cancers that arise at a single organ site, such as the ovary, kidney, or lung, represent a hodgepodge of different diseases at the molecular level. Morphologic features often allow the particular cancer types to be distinguished to some degree from one another. Yet even for patients whose cancers have essentially identical gross and microscopic appearances and very similar clinical manifestations, there may be vast differences in outcome. In spite of this complexity, the development of all cancers, regardless of type, is likely to be critically dependent on the acquisition of certain phenotypic features that allow the cancer cells not only to grow in an unchecked fashion in their tissue of origin but also to gain the ability to disseminate into surrounding tissues and organs, lymphatics, and the bloodstream and ultimately to grow as metastatic lesions in distant sites in the body.¹¹ As is indicated in Figure 13-2, among the signature traits that are likely to be expressed in the majority, if not all, of cancers are the following: (1) an increased tendency to manifest a stem cell or progenitor-like phenotype; (2) an enhanced response to growth-promoting signals; (3) a relative resistance to growth-inhibiting cues; (4) an increased mutation rate to allow for the rapid generation of new variant daughter cells; (5) the ability to attract and support a new blood supply (angiogenesis); (6) the capacity to minimize an immune response and/or evade destruction by immune effector cells; (7) the capacity for essentially limitless cell division; (8) a failure to respect tissue boundaries, allowing for invasion into adjacent tissues and organs as well as blood vessels and lymphatics; and (9) the ability to grow in organ sites with microenvironments markedly different from the one where the cancer cells arose. The development of some traits is likely to be associated with certain stages of tumorigenesis (see Fig. 13-2), but acquisition of signature traits in cancers is more likely to show a preferred order than an invariant order. Furthermore, many of the signature traits of cancer cells represent complex biological capabilities (e.g., angiogenic activity, immune evasion/resistance, and metastatic competence). Therefore it is likely that substantial changes in many signaling pathways are needed for the cancer cell to manifest these traits.

Figure 13-2 Acquisition of signature traits in neoplastic cells during cancer progression. Depicted in the figure are representative stages in the development of a cancer, perhaps a typical epithelial cancer, such as those that typically arise in the lung, colon, breast, or prostate. The schema suggests that most advanced cancers arise via clonal selection from subclones present in earlier stage benign and localized lesions (e.g., carcinoma in situ and locally invasive carcinoma). Some of the properties of advanced cancer cells are depicted by the ability of the cells to enter the bloodstream and to seed and grow in distant organ sites, such as the liver (*bottom*). Nine signature traits of cancer cells are listed. The relative time at which neoplastic cells acquire some of the traits is uncertain, although it seems likely that some traits, such as the expression of a stem/progenitor cell phenotype or enhanced response to growth-promoting signals, may be acquired earlier in cancer development. Other traits, such as invasive capacity and/or metastatic competence, may be acquired later. Many signature traits of cancer cells are in fact complicated biological capabilities (e.g., angiogenic activity, immune evasion/resistance, and metastatic competence) and likely depend on defects in a number of different factors and signaling pathways.

An exhaustive cataloging of the observations associating specific gene defects to the altered phenotypic traits of cancer cells will not be offered here, in part because of space limitations and in part because of uncertainties about the significance of many of the linkages between single gene defects and cancer phenotypic traits. Nonetheless, some general concepts regarding the relationships among gene defects and cancer cell phenotype have emerged, and two of these concepts are offered here. First, although some of the specific mutations and major gene expression defects that are seen in cancer cells may contribute predominantly to a few or perhaps even only one of the signature traits of cancer cells previously listed, it seems likely that many of the gene defects and expression changes have been selected for in large part because they exert pleiotropic effects on the cancer cell phenotype. Second, for certain tumor types, it has been possible to gain some insights into the apparent order in which genetic and epigenetic changes might arise and contribute to cancer pathogenesis. The data suggest that defects in certain genes and signaling pathways might be strongly selected for at a certain point in cancer development and progression, perhaps in large part because the alterations allow the precancerous or cancerous cells to acquire certain critical phenotypic features. Gene defects that might be nearly uniformly present in early-stage lesions of one tumor type might preferentially arise in later-stage tumors in another organ site. The data suggest that cellular and tissue context have critical, albeit poorly understood, modifying effects on the specific genetic defects that give rise to neoplastic transformation, clonal outgrowth, and tumor progression.

Two arguably more speculative points concerning genetic and epigenetic defects and the altered phenotypic traits manifest in cancer will be briefly discussed here. One issue is that the oncogene and tumor suppressor gene mutations present in cancer cells are likely to be far less “plastic” and reversible than some of the epigenetic alterations (e.g., DNA methylation and/or chromatin modification changes) that might achieve the same net effect on a particular gene or signaling pathway. Indeed, the reversibility of most epigenetic changes might allow cancer cells to alter signaling pathways, gene expression patterns, and cell phenotypes in a much more malleable fashion than mutational events. This characteristic perhaps allows the cancer cells to elaborate a particular collection of signature phenotypic traits in a specific tissue or physiological context and to extinguish, enhance, and/or elaborate other signature traits in a different tissue context (e.g., a distant metastatic site) or in response to therapeutic agents. A second issue is that, given that considerable genetic and epigenetic heterogeneity may exist within a primary cancer and among or even within individual grossly visible metastatic lesions, it remains uncertain whether each individual cancer cell in a large primary tumor mass need manifest many or all of the signature phenotypic traits or whether the cancer cells can essentially complement one another, such that the cancer population as a whole manifests all key traits, but few if any individual’s cells within a large tumor mass need manifest all traits. Furthermore, the evolving of different functions of individual precancerous cells may speak to the possibility of cooperation between different cell types in the codevelopment of the cancer and later in the process of tumor progression/metastasis.

Genetic and Epigenetic Defects in Cancer Alter Signaling Pathways and Regulators of Transcription, Chromatin, and Genomic Integrity

Recurrent Mutational Targets in Cancer

As discussed previously, in genetic terms, oncogenic alleles have gain-of-function mutations. Oncogenic variant alleles that are present in cancer are generated from the normal counterpart proto-oncogenes by various mutational mechanisms, including point or localized mutations, gross chromosomal rearrangements, or gene amplification. Some representative oncogene mutations in human cancer are summarized in Table 13-1. From a brief review of the data in Table 13-1, some generalizations can be offered. First, the mutations affect proteins functioning in various compartments of the cell, including growth factor receptors, cytoplasmic signal transducers, and nuclear proteins, such as transcription factors. Second, although some oncogene mutations may be unique to cancers of a particular type, such as the specific chromosomal translocations and resultant fusion proteins that are seen in cancers of hematopoietic origin (e.g., the BCR-ABL translocation that is seen in chronic myelogenous leukemia and a subset of acute lymphoid leukemias and the PML-RARα translocation that is seen in acute promyelocytic leukemia), other mutations, such as those affecting the KRAS, β-catenin, and c-MYC genes, are found in a broad spectrum of different cancer types. Third, oncogene mutations in cancer are nearly always somatic, because only a very limited number of germline mutations in proto-oncogenes have been linked to cancer predisposition thus far. Fourth, some proto-oncogenes, such as KRAS or BCL2, are somatically altered in cancer by a single mutational mechanism, namely, point mutations in the KRAS gene and chromosomal translocations affecting the BCL2 gene. In contrast, other proto-oncogenes, such as c-MYC, may be activated by more than one mechanism in cancer, including chromosomal translocation and gene amplification. Both mutational mechanisms lead to increased levels of c-MYC transcripts and protein. In fact, specific missense mutations at threonine 58 of the c-MYC gene in some lymphomas may further enhance c-MYC protein levels by abrogating a phosphorylation-ubiquitination mechanism targeting c-MYC for proteosomal degradation.¹² Of course, enhanced c-MYC protein levels can result from alterations in c-MYC–specific microRNAs, methylation, and other noncoding regulatory elements functioning to regulate c-MYC transcription and translation.

Table 13-1

Selected Oncogene Mutations in Cancer

Gene	Activation Mechanism	Protein Properties	Tumor Types
KRAS	Point mutation	Signal transducer	Pancreatic, colorectal, lung (adeno), endometrial, other carcinomas
NRAS	Point mutation	Signal transducer	Myeloid leukemia, colorectal cancer
HRAS	Point mutation	Signal transducer	Bladder carcinoma
EGFR (ERBB)	Amplification, mutation	Growth factor (EGF) receptor	Gliomas, lung (non–small cell) carcinoma
NEU (HER2/ERBB2)	Amplification	Growth factor receptor	Breast, ovarian, gastric, other carcinomas
c-MYC	Chromosome translocation	Transcription factor	Burkitt lymphomas
	Amplification		Small cell lung carcinoma (SCLC); other carcinomas; glioblastoma
MYCN	Amplification	Transcription factor	Neuroblastoma, SCLC; glioblastoma
MYCL1	Amplification	Transcription factor	SCLC, ovarian carcinoma
BCL2	Chromosome translocation	Antiapoptosis protein	B-cell lymphoma (follicular type)
CCND1	Amplification	Cyclin D, cell cycle control	Breast and other carcinomas
	Chromosome translocation		B-cell lymphoma, parathyroid adenoma
BCR-ABL	Chromosome translocation	Chimeric nonreceptor tyrosine kinase	CML, ALL (T cell)
RET	Chromosome translocation	GDNF receptor tyrosine kinase	Thyroid cancer (papillary type)
	Point mutation		Thyroid cancer (medullary type: germline mutations)
CDK4	Amplification
	Point mutation	Cyclin-dependent kinase	Sarcoma, glioblastoma
MET	Point mutation	Hepatocyte growth factor (HGF) receptor	Renal carcinoma (papillary type: germline mutations)
SMO	Point mutations	Transmembrane signaling molecule in sonic hedgehog pathway	Basal cell skin cancer
CTNNB1 (β-CAT)	Point mutation, in-frame deletion	Transcriptional coactivator, links E-cadherin to cytoskeleton	Melanoma; colorectal, endometrial, ovarian, hepatocellular, and other carcinomas, hepatoblastoma, Wilms tumor
FGF4	Amplification	Growth factor (FGF-like)	Gastric carcinoma
PML-RARA	Chromosome translocation	Chimeric transcription factor	APL
TCF3-PBX1	Chromosome translocation	Chimeric transcription factor	Pre-B ALL
MDM2	Amplification	p53 binding protein	Sarcoma
GLI1	Amplification	Transcription factor	Sarcoma, glioma
TTG2	Chromosome translocation	Transcription factor	T-cell ALL
AKT2	Amplification	Signal transducer (serine/threonine kinase; downstream effector of PI3K)	Pancreatic and ovarian carcinoma
PIK3CA	Amplification	Catalytic subunit of PI3K	Ovarian carcinoma
AURKA	Amplification	Centrosome-associated kinase	Breast, colon, ovarian, and prostate carcinomas gliomas
TMPRSS2-ERG	Chromosome translocation	Transcription factor (ETS family)	Prostate cancer
TMPRSS2-ETV1
TMPRSS2-ETV4

ALL, Acute lymphocytic leukemia; APL, acute promyelocytic leukemia; CML, chronic myelogenous leukemia; EGF, epidermal growth factor; FGF, fibroblast growth factor; GDNF, glial-derived neurotrophic factor; HGF, hepatocyte growth factor; PI3K, phosphatidylinositol 3-kinase; SCLC, small cell carcinoma of the lung.

In contrast to the activating mutations that generate oncogenic alleles from proto-oncogenes, inactivation of the normal function of tumor suppressor genes is critical in tumorigenesis. Akin to the proto-oncogenes, the functions of tumor suppressor genes are diverse, and proteins that are encoded by these genes reside in practically all subcellular compartments (Table 13-2). Many tumor suppressor genes were identified by virtue of the fact that they are mutated in the germline of persons who are affected by a known Mendelian cancer syndrome or who at the very least display a markedly elevated risk of cancer. The link between a germline-inactivating mutation in a purported tumor suppressor gene and increased cancer predisposition provides very persuasive evidence of the functional significance of the gene in the cancer process. Nevertheless, for the vast majority of tumor suppressor genes, in terms of their magnitude, somatic inactivating mutations play a far more significant role in cancer development than do germline mutations.

Table 13-2

Selected Tumor Suppressor Gene Mutations in Hereditary Cancer Syndromes and Sporadic Cancers

Gene	Associated Inherited Cancer Syndrome	Cancers with Somatic Mutations	Presumed Function of Protein
RB1	Familial retinoblastoma	Retinoblastoma, osteosarcoma, SCLC, breast, prostate, bladder, pancreas, esophageal, others	Transcriptional regulator; E2F binding
TP53	Li-Fraumeni syndrome	Approximately 50% of all cancers (rare in some types, such as prostate carcinoma and neuroblastoma)	Transcription factor; regulates cell cycle and apoptosis
CDKN2A (p16^INK4a protein)	Familial melanoma, familial pancreatic carcinoma	Approximately 25%-30% of many different cancer types (e.g., breast, lung, pancreatic, bladder)	Cyclin-dependent kinase inhibitor (i.e., CDK4 and CDK6)
CDKN2A (ARF protein)	Familial melanoma (?)	Approximately 15% of many different cancer types	Regulates MDM2 protein stability and hence p53 stability; alternative reading frame of CDKN2A gene
APC	FAP coli, Gardner syndrome, Turcot syndrome	Colorectal carcinomas, desmoid tumors, hepatocellular carcinoma, breast (rare)	Regulates levels of β-catenin protein in the cytosol; binding to EB1 and microtubules
WT1	WAGR, Denys-Drash syndrome	Wilms tumor	Transcription factor
NF1	Neurofibromatosis type 1	Melanoma, neuroblastoma	p21ras-GTPase
NF2	Neurofibromatosis type 2	Schwannoma, meningioma, ependymoma	Juxtamembrane link to cytoskeleton at adherens junction
VHL	von Hippel–Lindau syndrome	Renal (clear cell type), hemangioblastoma	Regulator of protein stability
BRCA1	Inherited breast and ovarian cancer	Ovarian (~10%), rare in breast cancer	DNA repair; complexes with Rad 51 and BRCA2; transcriptional regulation
BRCA2	Inherited breast (both female and male), pancreatic cancer	Rare mutations in pancreatic, others (?)	DNA repair; complexes with Rad 51 and BRCA1
MEN1	Multiple endocrine neoplasia type 1	Parathyroid adenoma, pituitary adenoma	Nuclear protein; unknown function
		Endocrine tumors of the pancreas
PTCH1	Gorlin syndrome, hereditary basal cell carcinoma syndrome	Basal cell skin carcinoma, medulloblastoma	Transmembrane receptor for sonic hedgehog factor; negative regulator of smoothened protein
PTEN	Cowden syndrome; sporadic cases of juvenile polyposis syndrome	Glioma, breast, prostate, follicular thyroid carcinoma, head and neck squamous carcinoma; many others	Phosphoinositide 3-phosphatase; protein tyrosine phosphatase
SMAD4	Familial juvenile polyposis syndrome	Pancreatic (~50%), approximately 10%-15% of colorectal cancers, rare in others	Transcriptional factor in TGF-β signaling pathway
BMPR1A	Familial juvenile polyposis syndrome	Not known	Receptor for bone morphogenetic protein
MSH2, MLH1 PMS1, PMS2, MSH6	Hereditary nonpolyposis colorectal cancer	Colorectal, gastric, endometrial, ovarian	DNA mismatch repair
CDH1	Familial diffuse-type gastric cancer	Gastric (diffuse type), lobular breast carcinoma, rare in other types (e.g., ovarian)	E-cadherin cell-cell adhesion molecule
STK11 (LKB1)	Peutz-Jeghers syndrome	Lung adenocarcinoma; rare pancreas cancers; absent in most other cancers	Serine/threonine protein kinase
EXT1	Hereditary multiple exostoses	Osteochondroma	Glycosyltransferase; heparan sulfate chain elongation
EXT2	Hereditary multiple exostoses	Not known	Glycosyltransferase; heparan sulfate chain elongation
TSC1	Tuberous sclerosis	Bladder carcinoma; head and neck squamous cancer; hepatocellular carcinoma	Hamartin; binds tuberin (TSC2); regulates cell size by inhibiting TOR function and protein synthesis
TSC2	Tuberous sclerosis	Head and neck squamous cancer	Tuberin (see previous row regarding TSC1)

FAP, Familial adenomatous polyposis; GTPase, guanosine triphosphatase; SCLC, small cell carcinoma of the lung; TGF, transforming growth factor; TOR, target of rapamycin; WAGR, Wilms tumor, aniridia, genitourinary anomalies, and mental retardation.

Another important point to consider is that much of the attention for tumor suppressor genes has been focused on demonstrating that cancer cells carry biallelic inactivating mutations. Clearly, a diverse array of mechanisms can inactivate gene function, including nonsense, frameshift, and nonconservative missense mutations, as well as splicing mutations and gross deletions of the gene or even the chromosome region that contains the gene. In a number of cases, studies of the chromosomal mechanisms associated with tumor suppressor gene inactivation in cancer tissues, such as loss of the parental heterozygosity (i.e., loss of heterozygosity) that is present in normal tissues, have even been used to infer the existence of tumor suppressor genes, in particular, chromosomal regions, before the actual identification of the tumor suppressor gene of interest. The emphasis on defining biallelic inactivating mutations in tumor suppressor genes has been stimulated in large part by the Knudson hypothesis,2,13 which predicted that recessive genetic determinants played a critical role in retinoblastoma and many other cancers and that inactivation of both alleles of a tumor suppressor gene was needed to abrogate tumor suppressor gene activity. Nevertheless, as will be discussed in a bit more detail in the following sections, a variety of observations indicate that epigenetic (nonmutational) mechanisms might play a prominent role in inactivating tumor suppressor gene function in sporadic tumors. Furthermore, for certain tumor suppressor genes, inactivation of only one of the two alleles of a tumor suppressor gene might significantly impair cell growth regulation or programmed cell death. For example, a number of mutant p53 proteins that carry missense mutations likely potently interfere via dominant negative mechanisms with the wild-type p53 protein in the cell because p53 functions as a homotetrameric protein and all subunits must be wild-type for fully intact p53 function in transcriptional regulation.¹⁴ For other tumor suppressor proteins, such as the cyclin-dependent kinase inhibitory protein p27, reduction of protein levels to 50% of the levels present in normal cells might result in significant detrimental effects on the ability of the cell to appropriately regulate growth.¹⁵

Epigenetic Mechanisms of Proto-oncogene Activation and Tumor Suppressor Inactivation

The preceding discussion largely emphasized the role and significance of somatic and germline mutations in proto-oncogenes and tumor suppressor genes in cancer. The rationale for focusing on the role of DNA sequence alterations in cancer initiation and progression is based chiefly on the view that it is more straightforward to ascribe a causal role in cancer to specific mutations in the tumor cell genome than it is to attribute a causal role in cancer to apparent changes in the levels and/or patterns of gene expression or varied DNA methylation or chromatin modification changes in cancer cells. Potential difficulties in assigning a causal role in cancer to changes in the expression of even specific proto-oncogenes or tumor suppressor genes, let alone genes of uncertain functional significance in cancer, include questions about whether the apparent differences in gene expression reflect authentic changes in gene expression or simply the possibility that the cancer might have arisen from neoplastic transformation of a cell type that had differences in gene expression from the majority of normal cells in the tissue from which the cancer originated.

In spite of the fact that changes solely in the expression, but not the structure or sequence, of proto-oncogenes and tumor suppressor genes have been more difficult to implicate definitively in the cancer process, major progress in the cancer epigenetics area has been made. Arguably, the most compelling data for assigning a critical and likely causal role to changes in gene expression in the cancer process are for the genes that have already been well established in prior mutational and/or function studies to be tumor suppressor genes, particularly in the cancer type under study. Hence, the emphasis here will be placed on illustrating how epigenetic mechanisms have been assigned a causal role in silencing tumor suppressor genes in cancer. This emphasis is largely due to space limitations and not simply because of the absence of data implicating epigenetic mechanisms in proto-oncogene activation in cancer. Indeed, a priori, there is no reason why epigenetic mechanisms cannot lead to substantial increases in the expression of proto-oncogenes, with expression changes in some cancers akin to those seen in cancers with high copy amplification of the respective proto-oncogene. In some cases, epigenetic mechanisms are likely to lead to overexpression of certain proto-oncogenes in a variety of cancer types, such as for c-MYC,¹⁶ the epidermal growth factor receptor,¹⁷ and the aurora-2 kinase and closely related kinases.¹⁸

As is summarized in Table 13-2, somatic inactivation of selected tumor suppressor genes has been well established to result from mutational (genetic) mechanisms in many cancers. However, a robust and growing body of data supports the view that epigenetic mechanisms somatically inactivate selected tumor suppressor genes in certain cancer types^19,20 (Fig. 13-3). Although data are still emerging on the likely diverse transcriptional and chromatin remodeling mechanisms responsible for epigenetic silencing of particular tumor suppressor genes, many studies have demonstrated that increases in the methylation of CpG-rich sequences (CpG islands) in the regulatory regions (i.e., promoter/enhancer) of tumor suppressor genes are often linked to loss of tumor suppressor gene expression. For instance, whereas the VHL gene is inactivated by mutational mechanisms in roughly 80% of renal carcinomas of clear cell type, in the majority of the clear cell renal carcinomas in which specific VHL mutations cannot be detected, loss of VHL gene expression appears to be tightly linked to hypermethylation of the VHL promoter.¹⁹ For some other tumor suppressor genes, including the CDKN2A, BRCA1, and MLH1 genes, promoter hypermethylation has also been implicated as key mechanism of inactivation.²⁰ In fact, on the basis of studies of genes that display extensive CpG island methylation and decreased or absent gene expression in cancer cells of one type or another, it has been suggested that aberrant CpG methylation might play a very broad and important role in silencing tumor suppressor genes in the cancer process.^19,20

Figure 13-3 Knudson’s two-hit hypothesis revised. In brief, Knudson’s hypothesis predicted that both alleles of a tumor suppressor gene would need to be inactivated by germline and/or somatic mutations to elicit critical phenotypic alterations associated with cancer development. The revised version of Knudson’s two-hit hypothesis considers the possibility that tumor suppressor gene inactivation can result from either genetic (mutation) or epigenetic silencing events. The two functional alleles of a given tumor suppressor gene are indicated by the two purple boxes (*top*). The first inactivating event affecting one of the two tumor suppressor gene alleles could be either a mutation, such as the localized defect indicated by the yellow box (*left*), or transcriptional silencing associated with or caused by hypermethylation of CpG-rich sequences in the promoter/regulatory region (*right*). The inactivating event for the second tumor suppressor gene allele (i.e., “Hit #2”) could be a nondisjunction event resulting in loss of the chromosome containing the wild-type tumor suppressor gene allele (loss of heterozygosity, LOH) or epigenetic silencing. (Modified from Jones PA, Laird PW. Cancer epigenetics comes of age. Nat Genet 1999;21:163–7.)

Nevertheless, at this point, it might be reasonable to offer a few cautionary comments regarding the linkages between CpG island methylation and gene silencing of specific tumor suppressor genes. One issue to reflect on is the uncertainty about what fraction of the genes whose promoters show increased methylation in cancers actually function in vivo as tumor suppressor genes. Promoter hypermethylation and loss of gene expression might be best considered as potentially useful but insufficient criteria for establishing tumor suppressor gene function in the absence of other supporting data. For instance, additional supportive evidence of tumor suppressor gene function might include data indicating that the methylation status of a promoter is tightly linked to its expression in a large panel of primary cancer specimens and data showing that gene expression can be readily and fully restored by treatment of cancer cells with demethylating agents and/or other agents that affect chromatin functional state, such as histone deacetylase inhibitors. In addition, data showing that biallelic inactivation of the methylated gene occurs by mutational mechanisms (e.g., localized mutation and loss of heterozygosity) or a combination of mutational and epigenetic mechanisms in at least some cancers might also represent a potentially critical set of observations. Yet another caveat to be aware of is that because transcription factors can act to repress tumor suppressor gene expression, in some cases promoter hypermethylation could in some cases be more of a reflection rather than a proximate cause of tumor suppressor gene inactivation in cancer.

Mutations Affecting Epigenetic Regulators in Cancer Cells

Colorectal cancers were among the cancer types studied initially in great depth for consistent patterns of epigenetic alterations, and it was long known that whereas most colorectal cancers display modest to moderate global loss of 5-methylcytosine (5mC), certain subsets of colorectal cancer were identified that appear to preferentially inactivate many different genes by promoter hypermethylation—the so-called CpG island methylator phenotype (CIMP).²¹ The findings on CIMP-positive colorectal cancers suggested that unidentified alterations in the methylation/demethylation machinery might play a role in the development of colorectal and perhaps other cancer types.

Indeed, during the past few years, intriguing insights into factors and mechanisms regulating DNA methylation state in cancer have emerged. Nearly all DNA methylation in somatic mammalian cells occurs at the cytosine residue of CpG dinucleotides and is catalyzed by DNA methyltransferases (DNMTs). Whereas DNMT mutations have been suggested to play a role in several different cancer types, it is now clear that localized mutations in DNMT3A are present in roughly 20% to 25% of patients with acute myeloid leukemia (AML).²² Although the DNMT3A mutations are associated with poor prognosis in persons with AML, it remains unclear exactly if and how the mutations affect DNA methylation in persons with AML.²² Besides mutations in DNMT3A and potentially other DNMT genes, other gene defects linked to DNA methylation factors and methylation state have been uncovered. A family of Tet methylcytosine dioxygenase proteins (TET1, TET2, and TET3) was identified in the past few years that modify DNA by hydroxylating 5mC to 5-hydoxymethylcytosine (5hmC) and, as a consequence, result in loss of 5mC in cellular DNA. More significantly for cancer development, inactivation of the TET2 gene was found in upward of 30% of all myeloid malignancies, and the TET2-mutant myeloid neoplastic cells have reduced levels of 5hmC and display a 5mC hypermethylator phenotype, with resultant changes in chromatin state and gene expression patterns.²³ Remarkably, further connections to altered DNA methylation in cancer have been revealed though studies of the linkages between mutations in the isocitrate dehydrogenase genes IDH1 and IDH2 and the hypermethylator phenotype in myeloid neoplasms and some other tumors.²³ Specifically, a critical cofactor for proper TET2 function is alpha-ketoglutarate, and the IDH proteins normally generate alpha-ketoglutarate from isocitrate. The mutant IDH proteins found in several cancer types, including myeloid cancers and gliomas, generate 2-hydroxyglutarate instead, which renders the TET2 protein in affected cells nonfunctional, with a resultant hypermethylator phenotype similar to that seen in TET2-mutant cancers.^23,24 It is expected that further studies will clarify the key genetic and epigenetic mechanisms underlying CIMP status in colorectal and other cancer types where the TET and IDH proteins appear to be functionally intact.

Although the progress in defining mutations that can underlie changes in 5mC and 5hmC state and DNA methylation in cancer has been highly encouraging, truly dramatic progress has been achieved during the past few years in defining a broad array of mutations that target chromatin modifiers and chromatin readers in cancers of nearly all types.²⁵ The chromatin modifiers include proteins that transfer or remove acetyl or methyl modifications from histone tails, as well as factors that move nucleosomes along the DNA, remove the nucleosomes entirely, or change the histone composition of nucleosomes in specific regions of the genome. Some of the factors that “read” the chromatin modifications are also mutated in cancer.²⁵ Furthermore, recent work indicates that even one of the variant histones, namely H3.3, which is associated with actively expressed gene regions as well as telomeres and pericentromeric chromatin, is mutated in upwards of one third of pediatric glioblastomas.^25,26 Table 13-3 summarizes some of the chromatin regulators that are recurrently altered in cancer. The global and specific functional consequences on the cancer epigenome of the varied chromatin regulator lesions in cancers of various types remain to be determined, but the consequences on chromatin state and gene expression are expected to be significant and broad. The findings offer an initial outline of what likely represents a complex web of connections between alterations in the cancer genome and the cancer epigenome.

Table 13-3

Selected Chromatin Regulators and Modifiers Recurrently Mutated in Cancer

Protein		Tumor Types with Mutations
HISTONE ACETYLTRANSFEREASES
CBP		AML, ALL, DLBCL, TCC
p300		AML, ALL, DLBCL, TCC, CRC, BrCA, PanCA
MOZ		AML, MDS
MORF		AML, EndometCA
HISTONE ACETYLATION READERS
BRD1		ALL
PBRM1		RenalCA, BrCA
HISTONE METHYLTRANSFERASES
MLL1		AML, ALL, TCC
MLL2		Medulloblastoma, RenalCA, DLBCL, FL
MLL3		Medulloblastoma, TCC, BrCA
SETD2		RenalCA, BrCA
EZH2		DLBCL, MDS, BrCA, ProstateCA, ALL
HISTONE DEMETHYLASES
JARID1A		AML
JARID1C		RenalCA
UTX		AML, TCC, RenalCA, EsophagealCA, MM
HISTONE METHYLATION READERS
TRIM33		ThyroidCA
ING1		Melanoma, BrCA
ING4		HNSCC
MSH6		CRC
CHROMATIN REMODELERS
BRG1		LungCA, Medulloblastoma, BrCA, ProstateCA, PanCA, HNSCC
BRM		HNSCC
ARID1A		OvCA, EndometCA, BrCA, RenalCA, CRC, GastricCA, Medulloblastoma, TCC
ARID1B		BrCA
BAF60A		BrCA
ATRX		Glioblastoma, PanNET
DAXX		Glioblastoma, PanNET
HISTONES
H3.1		Glioblastoma
H3.3		Glioblastoma

ALL, Acute lymphoblastic leukemia; AML, acute myeloid leukemia; BrCA, breast carcinoma; CRC, colorectal carcinoma; DLBCL, diffuse large B-cell lymphoma; EndometCA, endometrial carcinoma; EsophagealCA, esophageal carcinoma; FL, follicular lymphoma; GastricCA, gastric carcinoma; HNSCC, head and neck squamous carcinoma; LungCA, lung cancer; MDS, myelodysplastic syndrome; MM, multiple myeloma; OvCA, ovarian carcinoma; PanCA, pancreatic carcinoma; PanNET, pancreatic neuroendocrine tumor; ProstateCA, prostate carcinoma; RenalCA, renal carcinoma; TCC, transitional cell carcinoma of the bladder; ThyroidCA, thyroid carcinoma.

Modified from the findings and literature cited in Dawson MA, Kouzarides T. Cancer epigenetics: from mechanism to therapy. Cell 2012;150:12–27.

Genetic and Epigenetic Alterations and Genomic Integrity

A characteristic feature seen in most cancers is genetic instability, including high rates of localized point mutations or small insertion/deletion mutations in a subset of cancers and a chromosomal instability in a large fraction of many different cancer types.²⁷ The chromosome instability includes numeric abnormalities (aneuploidy), as well as simple and complex structural abnormalities. Considerable progress has been made in defining some of the factors and mechanisms that contribute to genetic instability in human cancer.

Roughly 2% to 4% of colorectal cancers (CRCs) arise in persons with hereditary nonpolyposis colorectal cancer (HNPCC), a familial cancer syndrome attributable to germline mutation in one allele of a mismatch repair (MMR) gene (Table 13-2; e.g., MSH2, MLH1, and MSH6).²⁸ The CRCs arising in persons with HNPCC arise in large part as a result of the somatic inactivation of the remaining wild-type MMR allele in an affected individual with HNPCC. The cancer cells manifest a markedly increased frequency of point mutations and small insertion and deletion mutations due to defective MMR function, reflected by the so-called high-frequency microsatellite instability (MSI-H) phenotype.²⁸ In the other cancer types arising in individuals and families with HNPCC, such as ovarian, endometrial, and gastric cancer, similar MSI-H phenotypes are seen. In the setting of HNPCC, there would appear to be more rapid tumor progression from an initiated clone to frank malignancy as a result of an MMR-defective mutator phenotype. Beside the HNPCC CRCs, about 10% to 12% of apparently sporadic CRCs manifest the MSI-H phenotype, and epigenetic silencing of MLH1 by DNA hypermethylation of the promoter region and perhaps other chromatin modifications plays a critical role in the majority of these sporadic MSI-H CRC cases.²⁸

Based on the results of recent comprehensive DNA sequence analyses of human cancers, it seems that hundreds to thousands of localized mutations can be detected in most primary cancers, yet the specific factors and mechanisms responsible for the mutator phenotypes in most of these cancers remain to be defined.¹⁰ Possible mechanisms include mutations and alterations of DNA polymerases, including aberrant expression of alternative specialized DNA polymerases known as translesional polymerases.²⁹

Functional studies in cultured cancer cells and in mice provide quite a broad array of mechanisms that could contribute to numerous chromosomal abnormalities and the associated chromosomal instability (CIN) phenotype in cancer. Such chromosomal abnormalities include mitotic checkpoint defects, chromatin modification or condensation defects, and chromosome segregation defects due to centrosome abnormalities and other possibilities, such as altered sister chromatid cohesion.³⁰ For the most part, although mutations in a few genes that might contribute to chromosome instability have been reported, the CIN phenotype in the vast majority of human cancers remains unexplained. A recent article indicates that the X-chromosome gene STAG2, a gene encoding a subunit of a sister chromatid separation complex, is mutated in significant subsets of melanomas, glioblastomas, and Ewing sarcoma, and that inactivation of STAG2 can promote aneuploidy in cultured cells.³¹ The findings highlight the possibility that further work on STAG2, STAG2-interacting proteins, and the other proteins that function in regulating proper segregation of chromosomes at mitosis will uncover more insights into the molecular basis of the CIN phenotype.

Structural chromosomal alterations are common in many cancer types and presumably reflect the consequences of DNA double-strand breaks and nonhomologous end joining to generate chromosomal deletions, inversions, and translocations, with biological selection acting on the variant cells with the structural alterations. If the structural abnormality activates and/or inactivates a gene or genes that lead to more robust proliferation and survival properties in a given context, cells carrying a given structural alteration will be positively selected for during cancer development. The fact that certain structural alterations are recurrent in cancer offers strong support for this concept. For the most part, particular structural chromosomal alterations in cancer were thought to be relatively straightforward, resulting in recombination of noncontiguous sequences, sometimes accompanied by deletions. However, recent work indicates that chromotripsis—a dramatic shattering of a chromosome or chromosomes and recombination of the fragments—may occur in 2% to 3% of all cancers and that in certain tumors, such as osteosarcomas, it may occur in up to 25% of cases.³² Chromotripsis may represent another mutator mechanism, because it would allow multiple genetic abnormalities to arise and be selected for concurrently rather than sequentially.

Alterations in Cancer Target Conserved Signaling Pathways and Networks

As noted previously and summarized in Tables 13-1 and 13-2, the protein products of proto-oncogenes and tumor suppressor genes have been implicated in diverse cellular processes. In light of the potentially enormous level of complexity suggested by the vast and diverse array of gene defects in cancer, it is somewhat reassuring that some general concepts have emerged with respect to the means by which genetic and epigenetic alterations likely contribute to cancer initiation and progression. A principal overarching theme is that the protein products of oncogenes and tumor suppressor genes function in highly conserved signaling pathways and regulatory networks.

The network in which the pRb tumor suppressor protein (encoded by the RB1 gene) functions is one of the more intensively studied oncogene–tumor suppressor gene networks. The pRb protein regulates cell cycle progression, in large part via its ability to bind to E2F transcription factor proteins.33,34 In addition to its role as a cell-cycle regulator, the pRb protein has been implicated in regulation of cellular differentiation, survival, and even angiogenesis in certain settings.³⁴ The binding of pRb to E2F proteins allows pRb to silence expression of E2F-regulated or “target” genes, such as those needed for the DNA synthetic (S) phase of the cell cycle. The ability of the pRb protein to bind to E2F proteins and to function in transcriptional repression appears to be tightly linked to its phosphorylation status, with the hyperphosphorylated forms of pRb incapable of binding to and regulating E2F proteins. The cyclin D1 protein and its associated protein kinase, cyclin-dependent kinase 4 (CDK4), negatively regulate pRb by phosphorylating it. The p16^INK4a tumor suppressor protein is a critical inhibitor of the CDK4/cyclin D1 complex (Fig. 13-4) and can therefore prevent the inactivation of pRb. As is noted in Table 13-2, a subset of sporadic cancers of various types has inactivating mutations in the RB1 gene. In other cancers, pRb function appears to be critically compromised as a result of mutations in other components of the network or pathway. For example, in many cancers that lack RB1 mutations, inactivating mutations in the CDKN2A gene have been noted. In other cancers, including some breast cancers, gene amplification and overexpression of cyclin D1 is found. In yet others, such as some glioblastomas and sarcomas, amplification and overexpression of the CDK4 gene has been seen. The net effect of mutations in the pRb pathway, whether in RB1 itself or in other genes, such as CDKN2A, CCND1 (cyclin D1), or CDK4, is to inactivate pRb function and its ability to regulate expression of critical E2F target genes (see Fig. 13-4).

Figure 13-4 Recurrent gene defects in conserved signaling pathways in cancer. Three signaling pathways that are commonly affected by mutations in various cancers are shown. Selected interactions between components of the pRb (*left*), APC/β-catenin (*middle*), and p53 pathways (*right*) are shown. Tumor suppressor proteins are indicated with red symbols, oncogene products are indicated by green, and proteins that are not known to be affected by mutational or epigenetic defects in human cancer are indicated in yellow. Inhibitory interactions between proteins are indicated by perpendicular lines, and activating effects are indicated by arrows. Presumptive downstream genes whose expression is affected by the pathways are noted. *APC,* Adenomatous polyposis coli; *AXI1,* axis inhibition protein 1; *β-CAT,* β-catenin; *Cdk4,* cyclin-dependent kinase 4; *CYC D1,* cyclin D1; *DHFR,* dihydrofolate reductase; *DNA Polα,* DNA polymerase α; *GADD45,* growth arrest and DNA damage inducible gene 45; *GSK3β,* glycogen synthase 3β; *MMP-7,* matrix metalloproteinase 7; *p21^CIP1,* p21 CDK-interacting protein 1; *p53AIP1,* p53-regulated apoptosis-inducing protein 1; *RNR,* ribonucleotide reductase; *TCF-4,* T-cell factor-4; *TS,* thymidylatesynthase; *TSP1,* thrombospondin 1.

Studies of other proto-oncogenes and tumor suppressor genes have also supported the existence of conserved regulatory networks in which multiple different tumor suppressor gene and proto-oncogene protein products function. Although the adenomatous polyposis coli (APC) tumor suppressor protein may have roles in regulating various processes in the cell,³⁵ a key function of APC is to participate in a multiprotein complex that regulates the levels of the β-catenin protein in the cytoplasm and nucleus (see Fig. 13-4). Components of the multiprotein complex that regulates β-catenin include the APC protein, another tumor suppressor protein known as AXIN1, and a kinase known as glycogen synthase kinase 3β (GSK3β). Inactivation of APC or AXIN1 function in cancer cells appears to lead to an inability to phosphorylate β-catenin and hence target it for recognition and subsequent ubiquitination by the βTrCP ubiquitin ligase and ultimately its destruction by the proteasome.^35,36 As a result, cancer cells with APC or AXIN1 inactivation display increased levels of β-catenin in the cytoplasm and nucleus and essentially constitutive complexing of β-catenin with transcription factors of the T-cell factor (TCF) family, such as TCF4. When bound to TCF4, β-catenin can function as a transcriptional co-activator, and in cancers with APC inactivation, such as colorectal carcinomas, TCF transcriptional activity is clearly deregulated. In a subset of the colorectal carcinomas that lack APC inactivation and in a variety of other cancer types (see Table 13-1), activating (oncogenic) mutations in the β-catenin protein have been found.³⁶ These missense and in-frame deletion mutations affect key phosphorylation sites in the N-terminus of β-catenin, essentially rendering β-catenin resistant to regulation by the APC/AXIN/GSK3β complex. Hence, the mutant β-catenin protein accumulates in the cell and deregulates TCF transcription. Transcription of selected proto-oncogenes may be activated directly by the β-catenin/TCF complex, such as c-MYC.³⁷

Other tumor suppressor gene regulatory networks have been defined, including the p53/MDM2/p19^Arf pathway (see Fig. 13-4), the PTCH/SMO/GLI pathway, and the MSH2/MLH1/PMS2 DNA mismatch recognition and repair pathway. Similar to the situation for the pRb, APC/β-catenin, and p53 pathways, mutations in cancer cells not infrequently target the PTCH/SMO/GLI and MSH2/MLH1/PMS2 pathways, either activating an oncogene within the pathway (e.g., SMO or GLI1 for the PTCH/SMO/GLI pathway) or inactivating one of the key tumor suppressors (e.g., either MLH1 or MSH2 in the mismatch repair pathway).

Although a large collection of genetic and biochemical data support the proposed protein functions and interactions depicted in Figure 13-4, it seems likely that the situation in vivo is far more complex. For example, on the basis of the regulatory scheme outlined for the pRb pathway in Figure 13-4, it might appear that the phenotypic consequences of RB1 or CDKN2A inactivation are functionally equivalent. However, patients with germline mutations that inactivate RB1 are predisposed to retinoblastomas and osteosarcomas, whereas persons with germline defects in CDKN2A are predisposed predominantly to melanoma and pancreatic cancer.^38,39 Furthermore, whereas persons with germline mutations that affect RB1 or CDKN2A are predisposed to a rather limited spectrum of cancers, somatic defects in the pRb pathway (e.g., including mutations in RB1, CDKN2A, CCND1 [cyclin D1], and CDK4) are seen in the majority of a broad array of cancer types.^38,39 Unfortunately, at present, although no compelling mechanistic explanation exists for these observations, perhaps a general explanation can be offered. Specifically, the genetic pathways in which certain oncogenes and tumor suppressor genes function are not simply linear pathways as indicated schematically in Figure 13-4 but more likely represent much more complex and branched networks. The branches of the network may even vary considerably, depending on cell type and developmental context, although it seems reasonable to predict that the genes and the protein products recurrently affected by mutation in human cancer represent particularly critical hubs in the pathways and networks.

Noncoding RNAs in Cancer—microRNAs and Long Noncoding RNAs

It has been recognized during the past decade that the abundance of messenger RNA (mRNA) transcripts and the translation of mRNA transcripts into proteins is highly regulated by a novel class of short noncoding RNAs, the so-called microRNAs (miRNAs). The mature forms of miRNAs are 20 to 24 nucleotides in length and are generated by successive cleavages by the Drosha and Dicer nucleases from longer precursor transcripts that contain characteristic hairpin formations.⁴⁰ Recognition of target transcripts occurs principally by binding of the miRNAs in the 3′ untranslated regions (3′UTR). Depending on the degree of homology to their target sequence, miRNAs induce translational repression or cleavage of mRNAs. Roughly 1000 human miRNAs have been described, and each miRNA can target hundreds to perhaps even a thousand or more mRNAs.⁴⁰ This ability allows for substantial combinatorial complexity and functional redundancy, making the identification of specific functions of miRNAs and their involvement in oncogenic or tumor suppressive networks difficult. The fact that miRNAs act via base-pairing of the miRNA with the 3′UTR also implies that alterations of both miRNA sequence and miRNA target sequence might have functional consequences. Thus far, large-scale sequencing efforts have focused more attention on the coding sequence of genes, which means that a significant number of functionally relevant alterations in noncoding regions might have been missed.

Several lines of evidence make it likely that miRNAs do play a role in the development of cancers. For instance, global inhibition of miRNA production seems to facilitate the acquisition of a neoplastic phenotype in primary mouse cells, suggesting that the net effect of miRNAs collectively might be a tumor suppressive effect.⁴¹ The genomic location of many miRNAs map close to common chromosomal breakpoints in cancer. In most cases, however, it is still unclear whether the miRNAs are actually involved in conferring the selective advantage that is gained by these translocations or whether miRNAs for other reasons are localized in regions of high genomic fragility.⁴² Comprehensive analyses of miRNA expression patterns in human cancers have revealed that different cancer types have distinct miRNA expression patterns. In fact, in many instances, miRNA expression patterns might be more precise in determining the tissue of origin than mRNA expression profiles.⁴³ Similar to the situation with protein-coding genes, in most cases, it is unclear which expression changes are causative and which are a result of the development of the tumors.⁴⁴

Several miRNAs seem to play a role as part of classic tumor suppressive or oncogenic signal transduction pathways. The miRNA17-92 locus has been described as a direct transcriptional target of the c-MYC and E2F oncogenes.⁴⁵ This polycistron locus encodes for seven miRNAs and has been shown to be genomically amplified and overexpressed in some human B-cell lymphomas and lung cancers. When overexpressed, it can cooperate with c-MYC to accelerate lymphoma development in a murine model system. In addition, miRNA372 and miRNA373 have been implicated as oncogenes in the development of germ-cell tumors at least in part by inactivating the p53 tumor suppressor pathway.⁴⁶ Recently miRNA10a has been suggested to be a driving force behind the metastatic progression of breast cancer cell lines.⁴⁷

Members of the let-7 family of miRNAs have been proposed as tumor suppressor genes in lung cancer, supposedly in part by their ability to inhibit the translation of the KRAS and HMGA2 oncogenes.⁴⁸ The members of the miRNA34 family have recently been shown to be direct transcriptional targets of the p53 tumor suppressor gene and may play a significant role in mediating the downstream functional activity of p53 in some settings.⁴⁹

Long noncoding RNAs (lncRNAs) represent another important class of noncoding RNAs in cancer pathogenesis.⁵⁰ By convention, lncRNAs are defined as a group of RNAs greater than 200 nucleotides in length that do not appear to encode a protein. Given the range of possible variations on this theme, many subclasses of lncRNAs have been suggested, often based on their location and orientation relative to other transcribed genes near their chromosome position. There is much circumstantial evidence, largely based on expression data, to circumstantially implicate many different lncRNAs in cancer development and progression, and space does not permit a full review of these data.⁵⁰ For the most part, relatively little functional data are available at this point that definitively implicates particular lncRNAs in cancer development or progression. Given the current uncertainties about which lncRNAs have critical functional roles in cancer development, some general concepts will be highlighted here instead. lncRNAs may function in a variety of ways to regulate the expression of various cellular genes, such as by serving as a scaffold for chromatin regulatory factors that regulate genes near the locus of the lncRNA and/or genes at other chromosome locations. Other lncRNAs may contain various miRNA binding sites and may functionally compete with other coding region transcripts for binding to particular miRNAs, thus acting in some settings to quite dramatically modulate the levels and ability of endogenous miRNAs to regulate transcripts of certain proto-oncogenes and tumor suppressor genes. Other functions for lncRNAs in cancer remain to be uncovered and substantiated, but it is possible that given the many structures that RNA can adopt and the catalytic functions of certain RNAs, lncRNAs may have a quite broad array of biochemical functions in cancer.

Role of Tissue and Context Differences in the Contributions of Gene Defects to Cancer Cell Phenotype

The published literature on the potential contributions of gene defects to the altered phenotype of cancer cells has offered some suggestions about how to consider the role of the gene defects in cancer pathogenesis. For instance, terms such as gatekeeper and caretaker have been used to classify the contributions of genes to cancer development.⁵¹ “Gatekeeper” genes have been suggested to be genes that play particularly critical roles in regulating cell proliferation and inhibiting cancer development in certain tissues, such as the APC gene in colorectal cancer; the tumor suppressive function of the genes must be overcome for cancers to arise in a given tissue or organ site.⁵¹ “Caretakers” have been generally defined as genes that do not play direct roles in growth control but rather likely play important roles in a number of tissues in maintaining the fidelity of the genome via their role in DNA damage recognition and repair processes. The MLH1 and MSH2 MMR genes have been proposed to be representative caretaker genes,⁵¹ and some researchers have suggested that perhaps the BRCA1 and BRCA2 genes also represent caretaker genes.⁵²

The use of terms such as gatekeeper and caretaker might have some merit. However, as will be illustrated in the following discussion, given the apparently important role of “gatekeeper” gene defects in cancers that arise in various organ sites but the quite variable timing of “gatekeeper” gene defects in the natural history of one cancer type versus another, the term gatekeeper might be more confusing than illuminating. In the case of some presumed “caretaker” genes, the genes might not be playing the passive role in the cancer process that has been assigned to them. This view is based on three lines of argument: (1) the apparent tissue specificity of the tumors that arise in persons who harbor germline mutations in the “caretaker” genes, such as in persons who are affected by hereditary nonpolyposis colorectal cancer and germline MLH1 or MSH2 mutations;⁵³ (2) the likely variable time in cancer development at which “caretaker” mutations may arise from one tumor to the next, with sporadic colon tumors not usually manifesting MMR gene inactivation and microsatellite instability until the carcinoma stage, despite the fact that adenomas in persons with hereditary nonpolyposis colorectal cancer often show high-frequency microsatellite instability;^54,55 and (3) the evidence that “caretaker” genes might actually play key roles in regulating cell proliferation and promoting apoptosis in certain contexts.⁵⁶

In general, persons who harbor a germline mutation in a specific tumor suppressor gene or proto-oncogene are predisposed to a very limited spectrum of cancer types. This observation is puzzling for a couple of reasons. The majority of genes that are affected by germline mutations in specific inherited cancer syndromes are essentially ubiquitously expressed in adult tissues. Furthermore, for a number of the tumor suppressor genes, mutations are often found to inactivate the gene in a much broader collection of sporadic cancer types than the types that commonly arise in germline mutation carriers. Children who carry a germline mutation in the RB1 gene have a very elevated risk of the development of retinoblastoma and a more modest risk of the development of osteosarcoma but no dramatic increase in the risk of most common adult cancers. Yet somatic defects in RB1 have been found and are believed to be critical in the development of many different cancers, such as small cell lung carcinomas, in which the vast majority have pRb defects.39,57

Potential explanations exist for these puzzling observations. For instance, whereas pRb might have an essential role in regulating retinoblast cell proliferation and/or differentiation, in other tissues, such as lung or breast epithelial cells, pRb might have a redundant role in growth control, perhaps because of the contribution of pRb-related proteins, such as p107 and p130.³³ Under this scenario, RB1 inactivation in most cell types might not promote neoplastic growth unless other defects, such as those in pRb-related proteins, are also present. An alternative and perhaps equally likely possibility is that somatic inactivation of pRb might trigger apoptosis in many cell types unless other somatic gene defects have arisen previously and these other defects interfere with the cell’s ability to undergo apoptosis after disruption of RB1 function. Evidence that pRb inactivation can act in a context-dependent fashion to promote apoptosis versus neoplastic transformation has been offered.^58,59 The tissue specificity of cancers that are seen in persons who carry germline mutations in inherited cancer genes is not restricted to the case of pRb. Germline p53 mutations predispose primarily to osteosarcoma, soft tissue sarcoma, leukemia, brain tumors, and breast cancer in women, and CDKN2A germline mutations predispose primarily to melanoma and pancreatic cancer.³⁸ In spite of the relatively limited spectrum of cancer types that are seen in people who carry germline p53 or CDKN2A mutations, the p53 and p16^INK4a genes are very commonly altered in human cancer, with each of the genes being inactivated in upward of 35% to 50% of many different sporadic cancer types.

Some genes with prominent roles in the development of a variety of different cancer types are sometimes presumed to have essentially singular functions in the cancer process in spite of data that suggest otherwise. As an example, the E-cadherin protein plays an important role in cell-cell adhesion via the ability of its extracellular domain to form adhesive interactions with E-cadherin molecules on opposing cell surfaces and the ability of the E-cadherin cytoplasmic domain to link to the actin cortical cytoskeleton via interactions with catenin proteins at the plasma membrane.⁶⁰ Early functional studies have suggested that restoration of E-cadherin in cancer cells that had endogenous E-cadherin defects interfered with the invasive properties of cancer cells in selected in-vitro assays.⁶¹ Perhaps in large part because of these observations, the loss of E-cadherin expression in cancer has often been assigned a role in promoting invasive behavior in advanced cancer cells. Although loss of E-cadherin function might indeed contribute to invasive behavior in cancers arising in vivo, it is worth bearing in mind that defects in E-cadherin could in fact play a distinct role in altering cell growth very early in the neoplastic transformation process in some tumor types, such as the gastric carcinomas that arise in patients who carry germline E-cadherin mutations.⁶²

Three additional examples of context-dependent effects of mutations in cancer will be offered here, in hopes of highlighting some of the potential challenges in unambiguously defining certain genes as oncogene targets or tumor suppressor gene targets in cancer development. The p53 gene represents one such example. As previously indicated, the wild-type p53 protein is a transcriptional regulator that is mutated in roughly 50% of all cancers. A common theme for the majority of the mutations is loss of wild-type p53 function as a transcription regulator, due to the mutant p53 proteins’ loss of sequence-specific DNA binding activity and/or misfolding of the protein.¹⁴ However, more than 75% of the p53 mutations in human cancer lead to expression of a mutant p53 protein that may have oncogenic activity independent of its ability to dominantly interfere with wild-type p53.¹⁴ Although the gain-of-function properties of mutant p53 in cancer are clear, the mechanisms underlying the oncogenic activity of mutant p53 are complex and likely vary depending on the particular mutant p53 protein and cellular context.¹⁴ A second example is the NOTCH1 gene, which is rarely mutationally activated by translocations in T-cell acute lymphocytic leukemia (T-ALL) and much more commonly activated in T-ALL by localized mutations in a juxta-membrane domain that allows the mutant NOTCH1 protein to be activated in a ligand-independent manner or by mutations in a cytoplasmic domain that lead to increased stability of the cleaved cytoplasmic domain of NOTCH1.⁶³ In contrast to the gain-of-function mutations in NOTCH1 in T-ALL, sequence-based studies in head and neck squamous cell carcinomas (HNSCCs) indicated that NOTCH1 harbors various inactivating mutations, including nonsense and frameshift mutations, in roughly 10% to 17% of primary HNSCCs.⁶³ Based on the well-known function of NOTCH proteins to regulate cell fate choices during development and adult tissues, it is perhaps not surprising that NOTCH1 can function as an oncogene when activated in some contexts and as a tumor suppressor gene when inactivated in other contexts.⁶³ A third example of the divergent and context-dependent function of a single gene in cancer is offered by a consideration of EZH2. The EZH2 protein is the catalytic component in the so-called polycomb repressive complex 2, which functions a methyltransferase for lysine 27 (K27) of histone H3.²⁵ Based on gene expression data showing EZH2 overexpression in subsets of prostate and breast cancer that were more aggressive clinically, as well as functional data implicating EZH2 overexpression in invasive and tumorigenic properties of prostate and breast cancers, EZH2 was implicated in cancer initially as an oncogene.⁶⁴ Other studies supported the view of EZH2 as an oncogene, because about 20% of diffuse large B-cell lymphomas had missense mutations at a single amino acid in the catalytic domain of EZH2 and the mutant EZH2 proteins had enhanced catalytic activity.²⁵ Contrasting with the data suggesting that EZH2 was uniformly an oncogene in human cancer are data indicating that EZH2 is acting by loss of function mutations in certain myeloid cancers, as well as T-ALL.²⁵

Clinical Implications

Even a far less than encyclopedic review of the genetic and epigenetic defects contributing to human cancer reveals significant clinical implications for the present and the future. Only a subset of some current directions and possibilities for the field are highlighted here. As previously noted, a very great and diverse array of mutations in oncogenes and tumor suppressor genes and epigenetic changes underlie different forms of cancer. However, because certain mutations and epigenetic changes either alone or collectively are preferentially and sometimes even uniquely associated with certain cancer types, molecular analyses of cancer specimens and biopsies are likely to be an important adjunctive approach in the cancer diagnostic area. For many hematologic malignancies and a subset of solid tumors, molecular analyses already serve a vital role in the diagnostic arena, helping, for instance, to distinguish subsets of tumors that manifest similar or overlapping histologies or to aid in diagnosis when the primary organ site of a metastatic cancer is uncertain. It seems likely that cancer diagnostic approaches that incorporate molecular annotation will be expand quite dramatically during the next 5 to 10 years. Moreover, based on recent reports that cell-free tumor-derived DNA can be readily detected in the plasma of patients with cancer and that copy number changes along with specific nucleotide changes can be detected with comprehensive sequencing approaches,65,66 it seems likely that molecular interrogation of plasma and/or other body fluids may well be pursued for improved staging information and to assess for relapse in patients with a solid tumor after surgery, radiation, and/or chemotherapy. Indeed, a recent early study suggests that molecular interrogation of circulating tumor-derived DNA in plasma may be useful for identifying drug-resistant cancer cells even prior to the initiation of certain therapies.⁶⁷

The genes and the protein products that are recurrently altered by mutations and/or epigenetic defects in human cancer more than likely represent particularly critical hubs in the pathways and networks that regulate cell growth, differentiation, and programmed cell death. Hence efforts to target the particular proteins and pathways with apparently central roles in the pathogenesis of certain cancer types would appear to offer potentially broad impact. Recent successes with BRAF inhibitors in the setting of BRAF-mutant melanomas ⁶⁸ and with ALK inhibitors in the setting of lung cancer harboring translocations activating ALK⁶⁹ offer evidence for the notion of targeting critical driver oncogene defects in cancer.

In spite of the generally optimistic view that is offered as a result of our rapidly expanding understanding of the nature and contribution of gene defects in cancer pathogenesis, some significant challenges remain if novel and specific new anticancer therapies are to be achieved in the near term. In particular, in a number of the cases with dramatic initial responses to single agents targeting the specific driver oncogene lesions, drug resistance and cancer recurrence and progression are seen. In certain other cancer types carrying some of the same driver oncogene lesions, such as BRAF-mutant colorectal cancers, the cancers show little if any evidence for response to the targeted agents.⁷⁰ Another issue is that for a fair number of the specific signaling pathways that are commonly disrupted in cancer, it might prove difficult to define readily tractable targets for therapeutic intervention. For instance, in the case of the p53 pathway, it is unclear how effectively p53 function can be restored for the proteins that carry missense substitutions that inactivate p53 tumor suppressor activity and perhaps also generate oncogenic p53 gain-of-function mutants. In the case of cancers with mutations leading to β-catenin deregulation, a potential goal might be to define agents that specifically interfere with the nuclear function of β-catenin in transcriptional activation of β-catenin/TCF-regulated target genes. Although this is not an unreasonable notion, given the rather limited successes to date in defining small molecules that specifically affect transcription factor complexes, it could be challenging to target β-catenin via conventional pharmacologic approaches. Similar concerns could perhaps be raised regarding the merits of attempting to specifically target other nuclear proteins and transcription factors that are deregulated in cancer cells.

Although the success of approaches to target cancer cells more selectively remains to be broadly established, it is obvious that ongoing and future efforts to understand the relationship between gene defects, altered cell signaling and physiology, and cancer phenotype have already shaped and will continue to shape our views of how best to proceed with novel therapeutic interventions. An optimistic view is that combining various targeted approaches with new molecular diagnostics efforts that inform about various issues—for example, disease subtypes, occult disease spread beyond the primary organ site, disease recurrence at earlier time points, and intratumor heterogeneity and resistance mechanisms—will yield major improvements in patient survival and quality of life for a range of cancer types within the next decade.