Cloning and analysis of the RB1 gene

The molecular cloning of the RB1 gene was facilitated by the identification in the chromosome 13q14 region of an anonymous DNA marker that detected DNA rearrangements in retinoblastomas.²⁶ Analysis of the DNA sequences flanking this DNA marker revealed a gene with the properties expected of RB1.^27–29 The RB1 gene has a complex organization, with 27 exons spanning more than 200 kb of DNA, and an RNA transcript of about 4.7 kb. The RB1 gene appears to be expressed ubiquitously rather than to be restricted to retinoblasts and osteoblasts.

Cloning of RB1 allowed study of mutations that inactivate the gene. Although gross deletions of RB1 sequences are observed in a small subset of retinoblastoma and osteosarcoma cases, most tumors appear to express full-length RB1 transcripts and lack detectable gene rearrangements when analyzed by Southern blotting.^30–33 Hence, the detection of inherited and somatic mutations in the RB1 gene in most cases requires detailed characterization of its sequence. Extensive analysis of mutant RB1 alleles has provided definitive molecular evidence supporting Knudson’s two-hit model. As predicted, patients with inherited retinoblastoma have been found to have one mutated and one normal allele in their constitutional (blood) cells. In retinoblastomas of such individuals, the remaining RB1 allele has been found to be inactivated by somatic mutation, usually by loss of the normal allele through a gross chromosomal event (Figure 2), but in some cases by point mutation. Multiple tumors arising in an individual patient with inherited retinoblastoma were all found to contain the same germline mutation, but different somatic mutations affected the remaining RB1 allele. The vast majority of patients with a single retinoblastoma and no family history of the disease have two somatic mutations in their tumors and two normal alleles in their constitutional cells.

The observation that RB1 is ubiquitously expressed is rather puzzling, given the restricted spectrum of tumors that develop in patients with germline RB1 mutations. Patients with germline mutations of RB1 are at elevated risk only for the development of a rather limited number of tumor types, including retinoblastoma in childhood, osteosarcoma, soft-tissue sarcoma, and melanoma later in life. RB1 germline mutations fail to predispose to most common cancers, despite the fact that somatic RB1 mutations have been observed in a wide variety of cancer types, including breast, small cell lung, bladder, pancreas, and prostate cancers.³⁴ It is possible that retinoblastoma protein functions differently in retinal epithelial cells than in other cell types, so that the RB1 gene acts as a “gatekeeper” in retinal cells but not in other cell types.

Function of retinoblastoma protein P105-RB

The protein product of the RB1 gene is a nuclear phosphoprotein known as p105-Rb or, more commonly, pRB. Its molecular weight is about 105,000 Da. Studies by Whyte and colleagues provided critical insights into pRB function, connecting human tumorigenesis with experimental tumors caused by DNA tumor viruses. They demonstrated that pRB formed a complex with the E1A oncoprotein encoded by the murine DNA tumor virus adenovirus type 5.³⁵ Prior studies of E1A had established that it had many effects on cell growth, including cell immortalization and cooperation with other oncogenes (e.g., mutated ras oncogene alleles) in neoplastic transformation. It was thus hypothesized that functional inactivation of pRB through its interaction with E1A might contribute to some of E1A’s transforming functions. Additional support for that proposal was provided by data establishing that mutations inactivating the ability of E1A to bind to pRB also inactivated E1A’s transforming function.^{36, 37} The significance of physical interaction between pRB and a DNA tumor virus oncoprotein was further supported by the subsequent demonstration that other DNA tumor virus oncoproteins also formed complexes with pRB, including SV40 T antigen and the E7 proteins of human papillomavirus (HPV) types 16 and 18 (Figure 3).^{38, 39} Many of the mutations that inactivated the transforming activities of these oncoproteins also inactivated their ability to interact with pRB. Furthermore, E7 proteins from “high-risk” HPVs (i.e., those linked to cancer development), such as HPV 16 and 18, formed complexes more tightly with pRB than did E7 proteins of “low-risk” viruses (e.g., HPV types 6 and 11). These studies of pRB provided compelling evidence that DNA tumor viruses might transform cells at least in part by inactivating tumor suppressor gene products. In addition, given the critical dependence of DNA tumor viruses on harnessing the cell’s machinery for replication of the viral genome, the studies also provided support for the hypothesis that pRB might control normal cell growth by interacting with cellular proteins that regulate the cell’s decision to enter into the DNA synthesis (S) phase of the cell cycle.

The functional activity of pRB is regulated by phosphorylation during normal progression through the cell cycle. Accordingly, pRB appears to be predominantly unphosphorylated or hypophosphorylated in the G1 phase of the cell cycle and maximally phosphorylated in G2 (Figure 4). The critical phosphorylation events regulating the function of pRB are likely to be mediated at the boundary between the G1 and S phases of the cell cycle by cyclin and cyclin-dependent kinase (Cdk) protein complexes.^{34, 40} When it is not phosphorylated, pRB forms complexes with proteins in the E2F family and inhibits transcription by recruiting proteins involved in transcriptional repression.⁴⁰ When phosphorylated, pRB can no longer efficiently form complexes with E2Fs (Figure 4). The E2F proteins, when dimerized with their differentiation-regulated transcription factor partner (DP) proteins, are then capable of activating the expression of a number of genes that are likely to regulate or promote entry into S-phase, including DNA polymerase a, thymidylate synthase, ribonucleotide reductase, cyclin E, and dihydrofolate reductase.⁴⁰ The E2F family members that directly affect cellular proliferation were shown in conditional mouse knockout models.⁴¹ Several other cellular proteins that bind to pRB have been identified, but their functions and the significance of their interactions with pRB remain less-well characterized than the interactions of pRB with E2Fs. Future studies will undoubtedly shed further light on the means by which loss of pRB function and that of pRB homologs p107 and p130⁴² contribute to cancer development.

TP53 gene

Studies in the late 1970s revealed that a cellular phosphoprotein with a relative molecular mass of about 53,000 Da formed a tight complex with SV40 T antigen; hence, the name of the p53 protein.^43–45 Further work established that TP53 also formed a complex with other viral oncogene products, including the adenovirus E1B protein, and that TP53 was present at low levels in normal cells and high levels in many tumors and tumor cell lines.^{43, 46–48} These initial findings suggested that increased levels of TP53 might contribute to cancer. Consistent with that notion, gene transfer studies provided data demonstrating that TP53 functioned as an oncogene in in vitro experiments.^{47, 49–51} However, subsequent studies in human tumors showed that TP53 was in reality a tumor suppressor gene.⁵²

The rationale for the human cancer studies was the observation that chromosome 17p LOH was common in a number of tumor types, including colorectal, bladder, breast, and lung cancer.^{53, 54} Detailed mapping showed that region 17p, which was lost in colorectal cancers, included the TP53 gene.⁵² Analysis of the sequence of the TP53 alleles retained in those cancers with 17p LOH demonstrated that the remaining TP53 allele was mutated,⁵² in perfect accord with Knudson’s hypothesis for the alterations expected in tumor suppressor genes. These observations were soon extended to other cancer types, and they explained many previous observations concerning TP53 that had been confusing when TP53 was believed to be an oncogene.^55–59 Additional evidence that TP53 functions as a tumor suppressor gene in human cancer is provided by gene transfer studies, but such overexpression studies cannot be easily interpreted, as many genes with no role in neoplasia can inhibit the growth of transfected cells.^60–63 Based on genome-wide sequencing of multiple tumor types, it is now clear that TP53 is believed to be the most frequently mutated gene in human cancers in general.⁶⁴

Germline mutations in the TP53 gene have been seen in those affected by Li–Fraumeni syndrome (LFS) and in a small subset of pediatric patients with sarcomas or osteosarcomas that do not meet the more strict criteria for diagnosis of LFS.^65–67 Those with LFS are at risk for developing a number of tumors, including soft-tissue sarcoma, osteosarcoma, brain tumors, breast cancer, and leukemia. Between one-half and two-thirds of patients with LFS have germline mutations in the central core domain of the TP53 coding sequences that resemble the somatic mutations frequently seen in sporadic cancers.⁶⁸ Some LFS patients and families with phenotypic features of LFS have germline mutations in a gene termed hCHK2 that phosphorylates TP53 and controls the cell’s response to DNA-damaging events.⁶⁹

In addition to somatic and inherited mutations in the gene, TP53 function can be inactivated by other mechanisms. As noted earlier, most cervical cancers contain high-risk or cancer-associated HPV genomes (i.e., HPV type 16 or 18). The E6 gene product of high-risk, but not low-risk, HPV types binds to a cellular protein known as E6-AP (for E6-associated protein) and stimulates TP53 degradation.^70–74 A cellular TP53-binding protein known as mouse double minute 2 (MDM2) is overexpressed in a subset of soft-tissue sarcomas as a result of gene amplification involving chromosome 12q sequences.⁷⁵ More recent studies have identified another TP53-binding protein, MDM4, which is overexpressed in a variety of cancers as a result of gene amplification of chromosome 1q sequences.⁷⁶ DNA transfection studies have shown that both the MDM2 and MDM4 genes can function as oncogenes when overexpressed. The oncogenic function is presumably mediated through their binding to and inactivation of TP53. Both proteins mask TP53’s transcriptional activation domain and promote TP53’s ubiquitination and subsequent degradation by the proteasome.^77–79 Consistent with the notion that MDM2 is a critical inhibitor of TP53 function, sarcomas with MDM2 amplification and overexpression rarely harbor somatic mutations in TP53.⁸⁰ Disruption of the MDM2 and MDM4 genes in the germline of mice is lethal, probably because such disruption allows unregulated activity of TP53. Accordingly, disruption of the murine TP53 gene rescues MDM2-deficient and MDM4-deficient mice from embryonic lethality.^{81, 82} Other mechanisms of regulating TP53 function have also been described, including mutation of a nuclear-cytoplasmic shuttle protein called nucleophosmin in almost 100% of adult acute myelogenous leukemias that demonstrate cytoplasmic localization of this protein, with the notable exception of acute promyelocytic leukemia.⁸³

TP53 function

The p53 protein has been shown to function as a transcriptional regulatory protein^{84, 85} though other, nontranscriptional functions for p53 have been described (ref. to recent review). In its wild-type state, the p53 protein is capable of binding to specific DNA sequences with its central core domain (Figure 5). The amino terminal sequences of p53 function as a transcriptional activation domain, and the carboxyl terminal sequences appear to be required for TP53 to form dimers and tetramers with itself. TP53 activates transcription of a number of genes with roles in the control of the cell cycle, including WAF1/CIP1/p21 (which encodes a regulator of Cdk activity),⁸⁶ MDM2 (as noted earlier, encoding a protein that is a known negative regulator of TP53), and 14-3-3 (a regulator of G2/M progression),⁸⁷ and various genes that likely function in apoptosis, including PUMA and NOXA. Experimental disruption of these genes by targeted homologous recombination can re-create some of the phenotypes associated with TP53 inactivation.^{88, 89}

The vast majority of the somatic mutations in TP53 are missense mutations leading to amino acid substitutions in the central portion of the protein (exons 5–9).⁹⁰ Consistent with the structure of the p53 protein, these missense mutations have marked effects on the p53 protein’s ability to bind to its cognate DNA recognition sequence through either of two mechanisms.⁹¹ Some mutations (e.g., mutations at codons 248 or 273) alter TP53 sequences that are directly responsible for sequence-specific DNA binding. Other mutations (e.g., codon 175) appear to affect the folding of TP53 and thus indirectly affect its ability to bind to DNA. Evidence that these missense mutations can confer “gain of function” rather than a dominant negative effect was demonstrated via “knock in” mouse models, whereby precise missense mutations were introduced into the endogenous TP53 gene.^92–96

Under some circumstances, TP53 acts at the G1/S checkpoint to regulate the cell’s decision to synthesize DNA, although TP53 also appears to have a critical function at G2/M.^{97, 98} In other settings, TP53 appears to exert control over the cell’s decision to undergo apoptosis or programmed cell death.⁸⁵ Of particular interest with regard to cancer treatment are data suggesting that some tumor cells lacking TP53 function are less sensitive to irradiation and chemotherapeutic agents such as cisplatin.^{92, 99, 100} However, studies of other tumor cells suggest that TP53 status shows a very different relationship to chemotherapeutic response, with cells that lack functional TP53 being markedly sensitive to DNA-damaging agents but resistant to 5-fluorouracil.¹⁰¹ Thus far, studies of primary human cancers have emphasized that a rather complex relationship is likely to exist between TP53 mutational status and the responsiveness of cancer cells to chemotherapy or radiation therapy, or both. In particular, it is difficult to distinguish the effects of TP53 mutation on the natural progression of disease from its effects on responses to treatment and other cellular stresses.^{102, 103} Hopefully, further research on TP53 will clarify our understanding of its normal functions, the basis for its frequent inactivation in many different cancers, and the consequences of its inactivation on tumor growth and response to therapy.

Cyclin-dependent kinase inhibitor 2A locus

Studies of the cyclin-dependent kinase inhibitor 2A (CDKN2A) locus on chromosome 9p illustrate well how observations from initially disparate lines of investigation often converge to implicate a particular locus as a critical factor in cancer development. The LOH of chromosome 9p was frequently found in many different tumor types, including melanoma, glioma, nonsmall cell lung, bladder, head and neck cancers, and leukemia.^104–107 Of considerable interest were observations establishing that a subset of such tumors had homozygous (complete) deletions affecting the 9p21 region,^108–110 strongly supporting the existence of a tumor suppressor gene in the region. In addition to the frequent somatic alterations of chromosome 9p sequences in cancers, linkage studies of some families with inherited melanoma indicated a melanoma predisposition gene mapped to essentially the same region of 9p.¹¹¹ These observations stimulated great interest in the chromosome 9p region presumed to contain the tumor suppressor gene or genes. One of the genes identified in the region as a result of positional cloning efforts was initially termed multiple tumor suppressor 1 (MTS1).¹¹² Sequence analysis of MTS1 showed that it was identical to a previously described gene that encoded the Cdk inhibitor protein known as p16.¹¹³ Because the p16 protein functioned by inhibiting Cdk4 and Cdk6, it was termed an inhibitor of cyclin-dependent kinase 4 (INK4) protein. Another highly related gene, mapping immediately next to the p16/MTS1 gene on chromosome 9p, was found to encode a second INK4 protein, known as p15 (Figure 6). The gene encoding the p16 protein is most often termed INK4A, and the gene for p15 is INK4B.^{114, 115} The approved symbols are CDKN2A and CDKN2B, respectively.

Subsequent studies show that heterozygous mutations in CDKN2A are present in some families with inherited predispositions to melanoma or pancreatic cancer.^116–119 Somatic mutations in CDKN2A are present in many different cancer types, including but not limited to melanoma, glioma, pancreatic and bladder cancers, and leukemia. In some tumors, deletions affecting the CDKN2A gene also involve the CDKN2B gene. In rare tumors, deletions inactivate CDKN2B but not CDKN2A.¹²⁰ The prevalence and specific nature of CDKN2A mutations vary markedly from one tumor type to another. In contrast to other tumor suppressor genes, such as RB1 and TP53, homozygous deletion is a fairly common mechanism of CDKN2A inactivation in cancer.¹²¹

Detailed studies of the CDKN2A locus led to the identification of a novel alternative transcript containing nucleotide sequences identical to those in transcripts for the p16^INK4A protein, but with unique 5′ sequences (Figure 6).^{114, 115, 122} The alternative CDKN2A locus transcript encodes a protein known as p14 alternative reading frame (p14^ARF). Of note, human p14^ARF is the same product as the originally identified mouse protein p19^ARF, which was designated as p19 due to its apparent weight.

The p14^ARF protein contains sequences from a distinct first exon (exon 1β). Exon 1β is located upstream of exon 1α, the first exon present in transcripts for p16 (Figure 6). Exon 1β is spliced to exon 2, which, along with exon 3, is present in the transcripts for both the p14^ARF and p16^INK4A proteins. However, the p14^ARF protein shares no sequence similarity with the p16^INK4A protein because p14^ARF synthesis initiates at a unique methionine codon in exon 1β and continues through exon 2, using an alternative open reading frame with no similarity to the p16^INK4A open reading frame. Careful studies of somatic and inherited mutations at the CDKN2A locus indicate that localized mutations inactivating the p16^INK4A protein are common in human cancer, with more than 60 germline mutations identified, but that localized mutations inactivating p14^ARF are uncommon.^{114, 115} However, the frequent occurrence of homozygous deletions at the CDKN2A locus implies that mutational inactivation of both proteins—and of p15^INK4B—may be strongly selected for during tumor development (Figure 6). Other findings suggest that p16^INK4A and p14^ARF expression may be lost in some tumor types as a result of methylation of DNA regulatory sequences at the CDKN2A locus (Figure 6).^123–125 Recent studies have demonstrated the importance of DNA methyltransferases in initiating and maintaining epigenetic silencing of the p16 tumor suppressor.^{126, 127} Furthermore, studies of mice with germline inactivation of p14^ARF and p16^INK4A indicate that these proteins function as tumor suppressor genes in vivo.^128–130

The mechanism through which the p16^INK4A protein controls tumorigenic growth is apparently through its inhibition of Cdk4 activity. As indicated earlier, phosphorylation of pRB impedes its ability to transcriptionally regulate E2F-target genes (Figure 4). The cyclin D1/Cdk4 complex has a critical role in regulating pRB phosphorylation and function.¹²⁴ Hence, the p16^INK4A protein, by virtue of its regulation of Cdk4 activity, is, in turn, a critical factor in regulating pRB phosphorylation. Presumably, inactivation of p16^INK4A results in inappropriate phosphorylation of pRB and a subsequent inability of hyperphosphorylated pRB to bind E2Fs and appropriately regulate gene expression at the G1–S transition. p14^ARF functions as a tumor suppressor by direct binding to the MDM2 protein, inhibiting the MDM2-induced degradation of TP53, thus maintaining appropriate function of TP53 in cells,¹¹⁵ much like p16INK4A maintains normal pRB function.

Contrary to this interpretation, however, is the fact that alterations of p14^ARF and TP53 often coexist in cancer cells, suggesting that they do not alter the same pathway, whereas Rb and p16INK4A alterations are mutually exclusive, supportive of the fact that they affect the same pathway.¹³¹ Nevertheless, these findings emphasize the concept that oncogenes and tumor suppressor genes do not function in isolation. Rather, they function in intricately linked cascades or networks that have important consequences for both tumorigenesis and therapy (Figure 7).^{85, 132}

Adenomatous polyposis coli gene

Identification and germline mutations

Colon cancers arising in a minority of hereditary syndromes do so from a state of polyposis. One of the polyposis syndromes is known as familial adenomatous polyposis (FAP) or adenomatous polyposis coli (APC). FAP is an autosomal dominant disorder affecting about 1 in 8000 individuals in the United States. The syndrome is characterized by the development of hundreds of adenomatous polyps in the colon and rectum of affected individuals by early adulthood. The lifetime risk of colorectal cancer in those with the classic form of FAP is extremely high, approaching 100% by age 60 years.

An interstitial deletion of chromosome 5q in a patient with features of FAP, but without any family history of the syndrome, greatly aided localization of the APC gene.¹³³ Subsequent DNA linkage studies confirmed that, in multiple kindreds with FAP, or the related condition known as Gardner syndrome, the polyposis phenotype segregated with DNA markers near 5q21.^{134, 135} In 1991, positional cloning efforts identified the APC gene as the specific gene responsible for FAP.^136–139 The APC gene is large, with more than 15 exons, and alternative splicing affects the 5′ untranslated portion of transcripts. The predominant APC transcript encodes a 2843-amino acid protein expressed in many adult tissues.

In the great majority of individuals with FAP, heterozygous germline mutations can be identified in the APC gene.^140–142 All of the germline APC mutations in those with FAP appear to inactivate APC protein function. The overwhelming majority of these germline mutations are localized nonsense or frameshift mutations in the 5′ half of the coding region of APC (Figure 8). Consistent with Knudson’s two-hit hypothesis, inactivation of the remaining wild-type APC allele by somatic mutation in those carrying a germline APC mutation is observed in the cancers that arise.^{143, 144} Correlations between the location of a particular germline APC mutation and clinical features have been found,¹⁴⁵ although clear insights into the molecular basis for the predisposition to extracolonic tumors (e.g., jaw osteomas and desmoid tumors) in FAP patients with variant phenotypes are lacking. However, some light has been shed on the variability in polyp number seen in families with polyposis.^{143, 146} Mutations in the 5′ region of the APC gene appear to be correlated with an attenuated phenotype attributable to reentry of the ribosome on the APC transcript downstream of the premature stop codon, resulting in an APC protein that retains some of its normal activity.¹⁴⁶ Mutations in 3′ third of the APC gene are also associated with a milder polyposis phenotype than mutations in the central third of the gene, perhaps because the mutated APC proteins similarly retain some tumor suppressor activity. Unexpectedly, extracolonic features such as desmoid tumors appear to be more common in patients with 3′ mutations.¹⁴³ Finally, a missense mutation in the middle of the APC gene has been found to predispose to colorectal cancers in Ashkenazi Jewish families.¹⁴⁷ This mutation does not alter the function of the gene product, but creates a “hot spot” that appears to be highly mutable, resulting in somatic deletions or insertions of surrounding nucleotides that produce truncations.

Function

The APC gene encodes a large protein of roughly 300 kDa that is hypothesized to regulate cell adhesion, cell migration, or apoptosis in the colonic crypt. The localization of the APC protein in the basolateral membrane of colonic epithelial cells, with an apparent increase in APC expression in cells near the top of the crypt, implies that APC may regulate shedding or apoptosis of cells as they reach the crypt apex.¹⁵¹ Perhaps consistent with this view, restoration of APC protein expression in colorectal cancer cells lacking endogenous APC expression has been reported to promote apoptosis.^{152, 153}

The APC protein binds to a number of proteins, including β-catenin, Y-catenin (also known as plakoglobin), glycogen synthase kinase 3β (GSK3β), end-binding protein 1 (EB1), human Drosophila large discs (hDLG), microtubules, and the related proteins axin and conductin.¹⁵⁴ With the exception of β-catenin, GSK3β, and the conductin and axin proteins, the significance and role of APC’s interactions with its various binding partners are not well understood. Several lines of evidence imply that APC has a critical function in regulating β-catenin.^{154, 155} β-Catenin is an abundant cellular protein, first identified because of its role in linking the cytoplasmic domain of the E-cadherin cell–cell adhesion molecule to the cortical actin cytoskeleton, via β-catenin’s binding to α-catenin. The truncated (mutant) APC proteins present in many colorectal cancers lack some or all of the repeat motifs crucial for binding to β-catenin. APC not only binds to β-catenin, but, in collaboration with the GSK3β enzyme and other proteins, such as axin and conductin, it appears to regulate the abundance of β-catenin in the cytosol via phosphorylation. In colorectal cancers in which APC is mutated and unable to bind or effectively coordinate the regulation of β-catenin, β-catenin accumulates in the cell, complexes with the transcription factor T-cell factor-4 (Tcf-4), and translocates to the nucleus (Figure 9). Once there, β-catenin functions as a transcriptional coactivator, activating expression of Tcf-4-regulated genes, typically bypassing regulatory proteins, such as Wnt. Consistent with the notion that β-catenin is a critical target of APC regulation, somatic mutations in β-catenin have been found in the small fraction of colorectal cancers lacking APC mutations.^156–158 These mutations consistently alter GSK3β phosphorylation consensus sites near the amino terminus of the β-catenin protein, and the mutations presumably render the defective β-catenin proteins oncogenic as a result of their resistance to degradation by APC and GSK3β. Although somatic mutations in APC appear to be rare in cancers arising outside the colon and rectum, oncogenic mutations in β-catenin’s N-terminus have been observed in many different cancer types.^{159, 160}

Wilms tumor gene

Wilms tumors are the most common renal neoplasm in children, accounting for approximately 6% of all pediatric cancers.¹⁶¹ Wilms tumors are similar to retinoblastomas in a number of ways: both can occur bilaterally or unilaterally, with single or multiple foci, and in a sporadic or inherited fashion. The two mutation model originally proposed for retinoblastoma was also proposed to explain Wilms tumor.¹⁶² However, hereditary cases are not as common among Wilms tumor patients as they are in retinoblastoma patients. Almost all patients inheriting a mutation at the RB1 locus develop a retinoblastoma, but only approximately 50% of individuals carrying a germline mutation predisposing to Wilms tumor develop the disease (i.e., lower penetrance).¹⁶¹

Perhaps the first finding to offer insight into an inherited genetic basis for Wilms tumor was a report describing six patients with Wilms tumor and sporadic aniridia (i.e., congenital absence of the iris).¹⁶³ It was proposed that the simultaneous occurrence of these two very rare conditions might result from chromosomal aberrations affecting two or more loci, a situation now often called a contiguous gene syndrome—mutation of one locus presumably leading to aniridia and mutation of another leading to Wilms tumor. This hypothesis was subsequently supported by the discovery of interstitial deletions of chromosome 11p13 in peripheral blood samples from children with the WAGR syndrome (Wilms tumor with aniridia, genitourinary abnormalities, and mental retardation) of Wilms tumor.¹⁶⁴ Cytogenetic studies of tumor tissues in a few cases of sporadic-type Wilms tumors revealed deletions or translocations of chromosome band 11p13.^{165, 166} Subsequent studies of paired samples of Wilms tumor and normal cells from patients, using probes that detect restriction fragment length polymorphisms (RFLPs) on chromosome 11p, revealed that LOH of 11p occurred frequently in Wilms tumors of both the inherited and sporadic types.^167–170

The Wilms tumor 1 (WT1) gene was identified in 1990 by virtue of mutations inactivating the gene in patients with the WAGR syndrome and by analysis of somatic mutations in the gene in tumors from a minority of patients with unilateral Wilms tumor and no associated congenital malformation.¹⁷¹ WT1 is encoded by 10 exons and its transcripts are subject to alternative splicing.^{172, 173} In contrast to the rather ubiquitous expression of the RB1, TP53, and APC genes, expression of the WT1 gene appears to be restricted to embryonic kidney and a small subset of other tissues.^{174, 175} WT1 messenger ribonucleic acids (mRNAs) encode proteins with molecular masses of 45,000–49,000 Da and 4 zinc-finger motifs. Based on their predicted amino acid sequences, the WT1 proteins were suspected from the outset to function in transcriptional regulation.¹⁷⁴ Several studies provide evidence to support that notion, although some WT1 isoforms may have a role in RNA processing, rather than in transcription regulation.^{173, 174} WT1 proteins suppress the transcriptional activity of promoter elements from a number of growth-inducing genes, including the genes for early growth response (EGR1), insulin-like growth factor-2 (IGF-2), and platelet-derived growth factor A (PDGFA) chain, suggesting that WT1 may function in gene repression.¹⁷⁶ Other studies suggest that WT1 can activate or repress gene expression, depending on the cell type and promoter context.¹⁷⁷ Consistent with the notion that WT1 may have a physiologic function in transcriptional activation, recent work indicates that WT1 activates expression of amphiregulin, a member of the epidermal growth factor family.¹⁷⁸ Loss of amphiregulin expression may contribute to loss of appropriate differentiation during Wilms tumor development. Adding to the complex nature of WT1’s role as a transcriptional regulator, recent studies suggest that certain WT1 splice variants have dramatically different effects in their ability to regulate gene expression,^{179, 180} including acting as an oncogene in a variety of adult cancers.

WT1 inactivation clearly contributes to Wilms tumor development in those with the WAGR syndrome. Moreover, approximately 10% of apparently sporadic Wilms tumors harbor detectable somatic mutations in the WT1 gene.¹⁸¹ Nevertheless, much evidence indicates that Wilms tumors arise through mutations in genes other than WT1. First, the chromosome 11p allelic losses seen in Wilms tumor frequently involve band 11p15, but not band 11p13, where the WT1 gene resides.^181–183 Second, the 11p15 region harbors a gene responsible for Beckwith–Wiedemann syndrome (BWS), a congenital disorder in which affected individuals manifest hyperplasia of the kidneys, endocrine pancreas, and other internal organs; macroglossia; and hemihypertrophy.^{184, 185}

Those affected by BWS are also at increased risk of developing embryonic tumors, such as hepatoblastoma and Wilms tumor. Finally, linkage studies of three families with dominant inheritance of Wilms tumor exclude linkage of the susceptibility locus in those families to any part of chromosome 11p.^{186, 187} These data and others suggest that germline mutations in any one of at least three different genes (i.e., WT1, the BWS gene, and at least one gene not on chromosome 11p reference to haber’s paper on X-chromosome gene) predispose to Wilms tumor. Whether a combination of inherited and somatic mutations in more than one of those genes is ultimately required for the transformation of a developing kidney cell into a Wilms tumor, or whether alternative genetic pathways for the development of Wilms tumors exist, remains to be established. The genetic heterogeneity observed among Wilms tumors provides an important contrast to the apparently less complex genetic pathway of retinoblastoma.

Neurofibromatosis 1 and 2 genes

von Hippel–Lindau gene

von Hippel–Lindau (VHL) syndrome is a rare dominant disorder predisposing affected individuals to the development of hemangioblastomas of the central nervous system and retina, renal carcinomas of clear cell type, and pheochromocytomas.^217–219 The VHL gene was mapped to chromosome 3p by linkage analysis. As with many other inherited cancer genes, LOH studies established that the VHL gene behaves as a typical tumor suppressor gene, with both alleles inactivated during tumorigenesis.^{218, 220} Positional cloning efforts identified the VHL gene in 1993.²²¹ Germline mutations inactivating one VHL allele are seen in the majority of individuals in families displaying features of the VHL syndrome.^217–219 As with some other inherited cancer syndromes, preliminary genotype–phenotype relationships have been observed. Specifically, a certain class of VHL germline mutations is associated with the development of renal cancer only, a second class is linked to predisposition to both renal cancer and pheochromocytoma, and yet a third mutation class is associated only with pheochromocytoma.²¹⁸ Somatic mutations in the VHL gene are also seen in more than 80% of sporadic renal cell carcinomas of the clear cell type, but not in renal cell carcinomas of other histopathologic types (e.g., papillary type).^{218, 219} Approximately, 20% of sporadic clear cell renal cancers do not carry a detectable mutation in the VHL gene. However, in many of these cases, the VHL gene may be inactivated by epigenetic silencing,²²² a mechanism noted earlier in this chapter in connection with inactivation of the CDKN2A locus. In tumor types other than clear cell renal cancer, inactivation of the VHL gene appears to be uncommon.²¹⁸

The VHL gene encodes a 213-amino acid protein whose major function appears to be in the regulation of angiogenesis through protein degradation. The protein encoded by VHL is part of a ubiquitin ligase complex that degrades hypoxia-inducing factor 1a (HIF-1a) in the presence of oxygen. In the absence of oxygen in normal cells, or when VHL is mutated in tumor cells, the HIF-1a transcription factor is stabilized, leading to the expression of cytokines such as vascular endothelial growth factor and the stimulation of angiogenesis. Further biochemical and cell biology studies on VHL and renal cell carcinomas are likely to offer definitive insights into angiogenesis, one of the most important stromal processes associated with neoplasia.^{223, 224}

Genomic stability genes

Several recessive cancer predisposition syndromes resulting from inactivation of genes that function in DNA damage recognition and repair have been described, including ataxia telangiectasia (AT), Bloom syndrome, xeroderma pigmentosum, and Fanconi anemia. In each case, the specific cancer type and the DNA-damaging agents that increase cancer risk are distinct. Although AT heterozygotes may subtly increase the risk of breast cancer,²²⁵ in other recessive cancer syndromes, only homozygotes appear to clearly increase cancer risk. This observation contrasts sharply with the picture in the dominant cancer predisposition syndromes discussed earlier (e.g., inherited retinoblastoma, FAP, NF1, and NF2), where heterozygotes have a clearly elevated cancer risk. Homozygous mutations in tumor suppressor genes rarely if ever exist, probably because the condition is lethal during embryogenesis. It is important to remember that the tumor suppressor genes do not only function as guardians against cancer; their main function is the control of normal cell balance, and their inactivation is expected to be incompatible with normal embryonic development.

Because recessive cancer syndromes are quite rare, this discussion of the role of genomic stability genes in cancer will focus on syndromes that are inherited in an autosomal dominant fashion. These syndromes include the most common types of familial cancers, predisposing to tumors of the colon, breast, and other organs.

DNA mismatch repair gene defects and hereditary nonpolyposis colorectal cancer

Familial clustering of colon cancer has long been recognized, with approximately 5% of all colon cancers attributable to inheritance of a gene defect with a strong effect on cancer risk and another 10–15% with a moderate effect on risk. Germline APC mutations are responsible for 0.5–1% of colorectal cancer cases in the Western world, and hereditary nonpolyposis colorectal cancer (HNPCC) is responsible for 2–4%.^226–228

Diagnostic criteria for identifying those individuals and families most likely to be affected by HNPCC have been determined,^226–229 despite the absence of overt clinical findings before cancer diagnosis, and the potential for chance clustering of colon cancer within a family. Representative diagnostic criteria include (1) exclusion of familial polyposis; (2) colorectal cancer in at least three relatives, one of them being a first-degree relative of the others; (3) two or more successive generations affected; and (4) at least one of the affected individuals being younger than 50 years of age at the time of diagnosis. Although not all individuals affected by HNPCC meet these criteria, familial aggregations of colorectal cancer that are likely to have a genetic basis distinct from that underlying the majority of HNPCC cases can be excluded. ^{226, 228}

Several genes responsible for HNPCC have been identified, including two on chromosome 2p [mutS homologs 2 and 6 (MSH2, MSH6)], another on chromosome 3p [mutS homolog 1 (MLH1)], and another on chromosome 7p [postmeiotic segregation 2 (PMS2)]. Together, germline mutations in these four genes account for virtually all classic HNPCC cases.^{226, 228–231} The protein products of the MSH2 and MLH1 genes appear to have critical roles in the recognition and repair of DNA mismatches (Figure 10). In cells with one normal and one mutant allele of a DNA mismatch repair (MMR) gene, DNA repair is minimally impaired, if at all. However, inactivation of the remaining allele can occur as a result of somatic mutation in a normal epithelial cell. This “second hit” abrogates MMR function, and hundreds to thousands of mutations may thereby occur during each subsequent cell-division cycle. Because these mutations preferentially arise in mononucleotide, dinucleotide, and trinucleotide repeat tracts (i.e., microsatellite sequence tracts), the phenotype is often called the microsatellite instability (MSI) phenotype.²²⁶

Germline mutations in the known MMR genes have been detected in only 2–4% of colorectal cancer patients, although approximately 15–20% of all colon cancers display the MSI phenotype.^{226, 229–234} Clearly, only a small fraction of the sporadic colorectal cancers with the MSI phenotype develop as the result of a germline mutation in a known MMR gene. Somatic mutations in MMR genes have been found in some sporadic colorectal cancers with the MSI phenotype.²³⁵ In most sporadic cases, however, inactivation of the MLH1 gene occurs as a result of epigenetic inactivation.^{236, 237}

Many of the mutations arising in cells with MMR deficiency are likely to be detrimental to cell growth. A small fraction of the total mutations that arise presumably activate oncogenes or inactivate tumor suppressor genes. Some genes are preferentially mutated in MMR-deficient cancers, presumably because the mutations confer a selective growth advantage. For instance, genes that contain repetitive DNA sequences, such as microsatellite tracts, might be expected to be targets of mutation in these cancers, and data support this prediction.

BRCA1 and BRCA2 genes

As in the case of colorectal cancers, family history has long been hypothesized to be a major risk factor in breast cancer. The greatest risk is seen in those who have a history of breast cancer in multiple first-degree relatives. However, only in the late 1980s was evidence obtained that predisposition to breast cancer in some families could be attributed to a highly penetrant autosomal dominant allele. In 1990, Hall and colleagues reported the localization of one such breast cancer predisposition gene, BRCA1, on chromosome 17q21.²³⁸ Others found that germline BRCA1 mutations substantially increase the risk not only of breast cancer but also of ovarian cancer.^{239, 240} Intensive research efforts were focused on the region of chromosome 17q harboring BRCA1, and the gene was ultimately identified by positional cloning approaches in 1994.^{241, 242}

Studies of germline BRCA1 mutations in breast cancer patients have yielded important results. In studies of families with four or more cases of breast or ovarian cancer (or both) diagnosed before age 60 years, germline BRCA1 mutations were identified in nearly 50% of families studied.^243–245 In fact, germline BRCA1 mutations may account for cancer predisposition in roughly 75% of families who manifest both breast and ovarian cancer.^{243, 244} Many distinct germline BRCA1 mutations have been identified, although most of the mutations result in the synthesis of a truncated BRCA1 protein.^{243, 244} Although most germline BRCA1 mutations have been identified in only one or a few families, some mutations have been found recurrently. The 11 most common mutations account for about 45% of the total BRCA1 mutations observed.^{244, 245} In fact, the two most common mutations in BRCA1 (185delAG and 5382insC) account for approximately 10% of the total. Of note, the 185delAG frameshift mutation at codon 185 of BRCA1, involving a deletion of two bases (adenine and guanine), has been identified in more than 20 Jewish families with familial breast or ovarian cancer. Moreover, population surveys of Ashkenazi Jews, chosen without regard to a family history of cancer, indicate that approximately 1% carry the 185delAG mutation.^244–246 Based on studies of families with germline BRCA1 mutations, the lifetime risks of breast cancer and ovarian cancer in those carrying an inactivating mutation are estimated to be 85% and 50%, respectively.^{243, 244, 247} Whether specific germline BRCA1 mutations confer a greater risk of breast and ovarian cancer than do other mutations remains uncertain.

As with most tumor suppressor genes and their associated familial cancer syndromes, germline mutations of BRCA1 lead to the presence of a mutant allele in every cell of the body. Cancers then arise through inactivation of the second wild-type allele by the mechanisms outlined in Figure 2. In the case of BRCA1 (and BRCA2), LOH of the remaining wild-type allele is usually responsible for the second “hit” leading to inactivation of the BRCA1 gene. In addition, LOH of the BRCA1 locus was found in approximately 50% of unselected breast cancers and 65–80% of unselected ovarian cancers.^{244, 248} Because most breast and ovarian cancers are not associated with a hereditary predisposition (i.e., they are sporadic), these studies of unselected cancers led investigators to hypothesize that BRCA1 would play an important role in the development of sporadic breast and ovarian cancers. However, despite these LOH data, sporadic breast and ovarian cancers rarely harbor mutations in the BRCA1 gene, demonstrating that at least one wild-type allele is still present in most of these sporadic cancers.^{244, 248} Somewhat surprisingly, this finding would suggest that BRCA1 does not have a role in the genesis of the more common, nonfamilial forms of breast and ovarian cancers. However, it is still possible that downstream effectors of the BRCA1 protein are altered in sporadic breast and ovarian cancers, suggesting that the pathway, rather than the gene, is an important contributor to the carcinogenic process of these cancers.²⁴⁹

Although germline mutations in the BRCA1 gene underlie cancer predisposition in roughly 40–50% of families with multiple breast cancer cases, another highly penetrant autosomal dominant susceptibility gene termed BRCA2 plays a critical role in a significant fraction of the families with multiple breast cancer cases lacking BRCA1 mutations. The BRCA2 gene was mapped to chromosome 13q12-13 in 1994²⁵⁰ and identified by positional cloning strategies in 1995.²⁵¹ Although germline mutations in BRCA1 and BRCA2 appear to confer essentially similar lifetime risks of female breast cancer (i.e., ∼80%), the risk of ovarian cancer is reduced to approximately 10% in those with BRCA2 mutations compared with approximately 40–50% in those with BRCA1 mutations. The risk of male breast cancer is markedly elevated in BRCA2 mutation carriers, with a lifetime risk of approximately 7%, as opposed to a 1% lifetime risk of in BRCA1 mutation carriers.²⁵² There also appears to be an elevated risk of pancreatic and several other cancers in both male and female BRCA2 mutation carriers.²⁴⁴ As is the case for BRCA1, LOH of the BRCA2 locus at 13q12, but not at the RB1 locus at 13q14, has been observed in some sporadic breast, pancreatic, head and neck, and other cancers, suggesting that BRCA2 may be a target for somatic mutations in cancer. However, as with BRCA1, detections of somatic BRCA2 mutations in sporadic cancers have been few,²⁴⁴ once again suggesting that perhaps the pathway, rather than the gene, is the target of genetic alteration in sporadic forms of these cancers.

The BRCA1 and BRCA2 genes each encode a large nuclear protein. The amino acid sequences of the two proteins have only short regions of similarity with one another or other well-characterized proteins. Although their lack of obvious functional motifs stymied initial attempts to define the cellular functions of BRCA1 and BRCA2, several lines of evidence indicate that both proteins interact directly or indirectly with homologs of yeast Rad51, a protein that functions in the repair of double-stranded DNA breaks.^253–260 Moreover, the BRCA1, BRCA2, and Rad51 proteins all appear to be present in a stable multiprotein complex in the cell’s nucleus. Consequently, it has been suggested that BRCA1 and BRCA2 may function in the response to or repair of DNA damage, particularly double-strand DNA breaks.²⁶¹ Other findings imply that BRCA1 and perhaps BRCA2 may have a role in regulating transcription.²⁶² Although the DNA repair and transcription regulation functions may be distinct, it is entirely possible that the two functions are linked in a process sometimes called transcription-coupled DNA repair.²⁵⁵ There is precedent for this idea, in that some nucleotide excision repair genes causing xeroderma pigmentosum can also function in transcription,²⁶³ and repair of certain types of DNA damage is known to be coupled to transcription.²⁶⁴

Candidate tumor suppressor genes

The tumor suppressor genes discussed earlier and others summarized in Table 1 are distinguished by the fact that germline-inactivating mutations in the genes are associated with inherited cancer predisposition. The link between germline mutation and elevated cancer risk provides incontrovertible evidence of the gene’s role in tumorigenesis. Other findings, such as the demonstration in sporadic cancers of LOH of one tumor suppressor gene allele accompanied by somatic mutation of the remaining allele, offer evidence for a more widespread role for many of the inherited cancer genes. Although the tumor suppressor genes in Table 1 are definitively linked to inherited cancer syndromes, it is possible that germline mutations in other bona fide tumor suppressor genes may be associated with a more modest cancer risk, as has already been demonstrated for the I1307K mutation in APC. A complete listing of all documented tumor suppressor genes is beyond the scope of this chapter and is ever-evolving as new data come to light from detailed sequencing studies of human cancers. Based on the prevalence of inactivating mutations identified in such studies, it is estimated that there are approximately 100 tumor suppressor genes that play a role in human cancers of various types.⁶⁴ The majority of these genes are only inactivated somatically and do not give rise to inherited cancer predisposition syndromes.

^a Representative genes of all major pathways and hereditary cancer predisposition types are listed. Approved gene symbols are provided for each entry; alternative names are in parentheses.

^b In many cases, the gene has been implicated in several pathways. The single pathway that is listed for each gene represents a “best guess” (when one can be made) and should not be regarded as conclusive.

^c In most cases, the nonfamilial tumor spectrum caused by somatic mutations of the gene includes those occurring in familial cases but also additional tumor types. For example, mutations of TP53 and CDKN2A are found in many more tumor types than those to which Li–Fraumeni and familial malignant melanoma patients, respectively, are predisposed.

Abbreviations: APC, adenomatous polyposis coli; APOP, apoptotic pathway; BER, base excision repair; CIN, chromosomal instability; FAP, familial adenomatous polyposis; GLI, glioma-associated oncogene; HIF1, hypoxia-inducing factor 1; MMR, mismatch repair; NER, nucleotide excision repair; PI3K, phosphatidylinositol 3-kinase; RB, retinoblastoma; RTK, receptor tyrosine kinase pathway; SMAD, SMA- and MAD-related protein 4; wt, wild type.

Source: Adapted from Vogelstein and Kinzler 2004.¹³¹ Reprinted by permission from Macmillan Publishers Ltd.

Studies of RNA have revealed a large number of genes whose transcription is reduced in cancers. Some of these genes were initially termed tumor suppressors solely on the basis of this reduced expression. A small subset of these genes may, indeed, have critical roles in growth regulation, but expression data provide insufficient evidence to invoke causality. The altered expression of genes in cancers more often reflects the effect rather than the cause of the neoplastic process (i.e., the altered growth and differentiation properties of cancer cells and their abnormal microenvironment compared with normal cells in the tissue or organ from which the cancer arose). Moreover, many genes that play no role in cancer can dramatically alter cell growth when expressed exogenously at high, unphysiologic levels, rendering functional analyses of such candidates problematic. In the end, the functional evidence should be carefully weighed before a conclusion can be drawn that a gene has a causal role in tumorigenesis and that it should appropriately be designated a tumor suppressor gene. The only definitive way to identify a tumor suppressor gene at present is to document a highly statistically significant number of inactivating mutations in it in specific cancer types.⁶⁴

Summary

There is now overwhelming evidence that mutations in tumor suppressor genes are major molecular determinants for most common human cancers. Thus far, dozens of confirmed tumor suppressor genes have been identified by molecular cloning techniques. In some cases, these genes are inactivated in the germline, and their inactivation predisposes to cancer. Far more frequently, these same tumor suppressor genes are inactivated by somatic mutations during tumor development.

Although much has been learned about tumor suppressor genes, a great deal of work remains. Advancements in our understanding of tumorigenesis will continue with the detailed characterization of their normal cellular functions and the mechanisms through which mutations of these genes subvert these normal functions. It is essential to point out that tumor suppressor genes, unlike oncogenes, are in general not directly targetable by drugs: all drugs inactivate their target proteins and tumor suppressor genes are already inactivated in cancers. Therefore, the best hope for exploiting tumor suppressor gene mutations in cancers is through targeting downstream pathways that are consequently upregulated. Understanding the pathways through which these genes act is therefore of paramount importance.

---

Only gold members can continue reading. Log In or Register to continue

Share this:

• Click to share on Twitter (Opens in new window)
• Click to share on Facebook (Opens in new window)

Related

Related posts:

Aberrant signaling pathways in cancer Disparities in cancer care Monoclonal serotherapy Neoplasms in acquired immunodeficiency syndrome Neoplasms of the small intestine, vermiform appendix, and peritoneum and carcinoma of the colon and rectum Infections in patients with cancer

Stay updated, free articles. Join our Telegram channel

Join