Cell, Vol.

116, 235–246, January 23, 2004, Copyright 2004 by Cell Press

Principles of Tumor Suppression Review

Charles J. Sherr erties of “classic” tumor suppressor genes. First, they

Howard Hughes Medical Institute are recessive and undergo biallellic inactivation in tu-
Department of Genetics and Tumor Cell Biology mors. Second, inheritance of a single mutant allele ac-
St. Jude Children’s Research Hospital celerates tumor susceptibility, because only one addi-
332 North Lauderdale tional mutation is required for complete loss of gene
Memphis, Tennessee 38105 function. Hence, a germline mutation can be the underly-
ing cause of a familial cancer syndrome that exhibits an
autosomally dominant pattern of inheritance. Third, the
same gene is frequently inactivated in sporadic cancers.
Beyond this, strict constructionists might argue that a
Molecular genetic studies of familial cancer syn- tumor suppressor is a gene that, when restored to activ-
dromes identified and defined the recessive nature of ity, can reverse the tumorigenic properties of a cell, a
tumor suppressor genes and resolved the paradox requirement that rarely has been met experimentally.
of why tumors arising in such families exhibited an Indeed, given that tumor suppressors can prevent the
autosomally dominant pattern of inheritance. Subse- acquisition of additional deleterious mutations that might
quent characterization of tumor suppressor proteins otherwise provide cancer cells with a further selective
revealed their widespread involvement in sporadic advantage, tumor suppressor gene inactivation might
cancers and pinpointed key mechanisms that protect well allow further genetic changes that could confer
animals against tumor development. We now recog- resistance to their restoration at a later time.
nize that tumor suppressor genes regulate diverse cel- Because many others have reviewed the properties
lular activities, including cell cycle checkpoint re- of individual tumor suppressor genes in detail, I focus
sponses, detection and repair of DNA damage, protein here on selected members in order to exemplify the
ubiquitination and degradation, mitogenic signaling, scope of their biologic activities. I then touch briefly on
cell specification, differentiation and migration, and the concepts of haploinsufficiency (“one-hit” inactiva-
tumor angiogenesis. Their study has become a center- tion) and combinatorial interactions between genes that
piece of contemporary cancer research. more subtly provide tumor resistance. Due to limitations
of space, I have chosen to emphasize the roles of repre-
It has long been recognized that cancers arise as a result sentative tumor suppressor genes in human cancer,
of somatic mutations, a concept dramatically reinforced making less frequent references to mouse model systems.
by the demonstration that cellular “protooncogenes,”
when mutationally deregulated or abnormally overex- The Checkered History of p53
pressed, contribute to tumor formation (Stehelin et al., The path to discovery of the p53 tumor suppressor—the
1976). Our understanding that many such genes encode starting point for this review (Finlay et al., 1989)—was
proteins that govern processes of cell proliferation, dif- filled with more than a few twists and turns. Linzer and
ferentiation, and development, and that mutations af- Levine (1979) and Lane and Crawford (1979) first de-
fecting their functions constitutively deregulate particu- tected p53 in complexes with SV40 T antigen. Oren and
lar signaling pathways provided us with some of the Levine (1983) cloned p53 cDNA from SV40-transformed
clearest mechanistic insights about how and why cancer cells and reported, as did Robert Weinberg’s laboratory,
cells misbehave (Bishop and Varmus, 1985). The discov- that p53 could collaborate with mutant Ras to transform
ery of genetically dominant, “activated” oncogenes also primary rat embryo fibroblasts (Eliyahu et al., 1984; Par-
fueled the idea that a distinct class of “antioncogenes” ada et al., 1984). On this basis, it was reasonably con-
might oppose their effects and block tumor develop- cluded in the parlance of the time that p53 was an “im-
ment. Indeed, experiments involving somatic cell fusion mortalizing oncogene,” a protein that, like Myc or
and chromosome segregation had pointed to the exis- adenovirus E1A, could collaborate with Ras to transform
tence of genes that could suppress tumorigenicity (Har- primary rodent cells.
ris et al., 1969; Stanbridge, 1976). Over the last 15 years, As a harbinger of what was to come, others found that
many such tumor suppressor genes have been identified the cellular p53 gene was rearranged and inactivated in
(Table 1). Because their cancer-preventive effects usu- mouse erythroleukemia cells by insertion of the Friend
ally require the presence of only a single functional gene, murine leukemia virus into the locus (Mowat et al., 1985).
prototypic tumor suppressor genes are recessive, re- These changes were observed to occur in vivo during
quiring “two-hit” inactivation of both alleles (Knudson, the natural course of virus-induced disease, although
1971, 1973; Comings, 1973). Thus, the earliest attempts the precise nature of the selective advantage conferred
to identify them relied on genetic approaches that fin- by p53 disruption remained unclear. Things began to
gered instances of biallelic gene inactivation, typically further unravel when a murine p53 cDNA derived from
in a setting in which one mutated allele was inherited F9 embryonal carcinoma cells failed to collaborate with
through the germ line and the other was lost somatically. Ras in the cotransformation assay but formed foci of
In retrospect, these features define three cardinal prop- transformed cells when mutated (Hinds et al., 1987).
Ensuing investigations “call(ed) into question what the
Correspondence: [email protected] correct p53 wild-type sequence is and whether a wild-
Cell
236

Table 1. Representative Tumor Suppressor Genes

Gene Function Familial Cancer Association Other Major Tumor Types

p53 Transcription factor Li-Fraumeni syndrome ⬎50% of cancers

RB Transcriptional corepression Retinoblastoma Many
INK4a (p16) Cdk inhibitor (RB activation) Melanoma Many
ARF Mdm2 antagonist (p53 Melanoma Many
activation)
APC Wnt/Wingless signaling Familial adenomatous Colorectal cancer
polyposis
PTCH Hedgehog signaling (receptor) Basal cell nevus (Gorlin) Medulloblastoma, basal cell carcinoma,
syndrome rhabdomyosarcoma
SMAD4/DPC4 TGF-␤ signaling (Transcription Juvenile polyposis Pancreatic and colon cancer
factor) (hamartomas)
PTEN Lipid phosphatase Cowden syndrome Glioblastoma, endometrial, thyroid, and
(phosphoinositide prostate cancers
metabolism)
TSC1,2 GTPase activating protein Tuberous sclerosis Renal cell carcinoma (rare),
complex (mTOR inhibition) (hamartomas) angiofibromas
NF1 GTPase activating protein for Neurofibromatosis Sarcomas, gliomas
Ras
WT1 Transcription factor Wilm’s tumor
MSH2 and MLH1 DNA mismatch repair Hereditary nonpolyposis Endometrial, gastric, ovarian, bladder
colorectal cancer cancer
(Lynch syndrome)
ATM DNA damage sensor (protein Ataxia telangiectasia Lymphoreticular malignancies
kinase) (T-cell lymphoma)
NBS1 DNA repair, S phase checkpoint Nijmegen breakage Lymphoreticular malignancies
control syndrome (T cell
lymphoma)
CHK2 Protein kinase (G1 checkpoint Li-Fraumeni syndrome
control)
BRCA1, BRCA2 DNA repair Familial breast and ovarian
cancer
FA genes DNA repair, S phase checkpoint Fanconi Anemia Acute myelogenous leukemia
VHL E3 ligase recognition factor for Von Hippel-Lindau Renal cell carcinoma, cerebellar
HIF␣ syndrome hemangiosarcoma

type p53 gene can transform cells in culture” (Finlay et and even oncogene activation (Prives, 1998) (Figure 1).
al., 1988). Moreover, the formation of oligomeric com- In turn, p53 orchestrates a global transcriptional re-
plexes between the wild-type and mutant p53 proteins sponse that either counters cell proliferation or, more
raised the possibility that the mutationally inactivated dramatically, induces apoptosis. Its reputation as a tu-
forms might act in a transdominant manner to inhibit the mor suppressor is secure, as p53 is now recognized to
function of the wild-type protein (as they can!) (Eliyahu et be the singly most frequently inactivated gene in human
al., 1988; Hinds et al., 1989). Finlay and coworkers (1989) cancers (Olivier et al., 2002).
concluded this chapter with their landmark paper dem-
onstrating that p53 can act as a suppressor of transfor- RB, the First “Classic” Tumor Suppressor
mation. Backing up a bit, the cardinal features of tumor suppres-
Recognizing that p53 was deleted in human colorectal sion were first exemplified in studies of retinoblastoma
cancers, analysis of the second p53 allele in tumor cells and Wilm’s tumor before p53 was identified. Alfred
showed that it had sustained mutations, implicating p53 Knudson’s perspective as a pediatrician and cancer ge-
loss as a driving force (Baker et al., 1990). Mutations of neticist sparked his interests in childhood malignancies
p53 were soon documented in many other forms of in which hereditary features were manifest. He articu-
sporadic cancer and were revealed to be a causative lated the idea that retinoblastoma might be caused by
genetic factor in patients with the familial Li-Fraumeni two mutations, one of which might be inherited through
cancer susceptibility syndrome (Malkin et al., 1990). In- the germ line (Knudson, 1971, 1973). Families transmit-
triguingly, the demonstration that p53 is a sequence- ting such mutations would manifest a pattern of domi-
specific DNA binding protein was made only later (Kern nant inheritance, so that affected children would de-
et al., 1991). The wild-type protein was soon revealed velop disease early in life that frequently affected both
to be induced by DNA damage and to cause G1 phase eyes. In nonhereditary cases requiring two de novo mu-
arrest, suggesting that p53 performs a cell cycle “check- tations, the disease would be rare, develop later, and
point” function that guards cells against genotoxic insult be unilateral almost without exception. In 1976, several
(Kastan et al., 1991). We now appreciate that p53 is a groups, Knudson’s among them, used banding tech-
homotetrameric transcription factor that is activated in niques to demonstrate interstitial deletions of chromo-
response to many forms of cellular stress, including some 13q14 in retinoblastoma, leading to speculation
irradiation, hypoxia, drug-induced genotoxic damage, that “the RB gene” might reside at this locus (Knudson
Review
237

metabolism and replication and whose expression is

required to enable cells to enter the DNA synthetic (S)
phase of the cell cycle (Nevins, 2001; Trimarchi and
Lees, 2002). When bound at E2F-responsive promoters,
RB family proteins help to repress gene expression by
recruiting histone deacetylases and chromatin-remod-
eling factors to these loci (Harbour and Dean, 2000).
By contrast, phosphorylation of RB family proteins by
mitogen-activated, cyclin-dependent kinases cancels
RB-mediated repression. Importantly, this provides a
signaling pathway linking extracellular cues to the mo-
lecular apparatus that controls the initiation of DNA repli-
cation in mammalian cells (Weinberg, 1995) (Figure 1).
Loss of RB weakens these controls, dissociating the
cell cycle machinery from extracellular signals, dampen-
ing the ability of proliferating cells to exit the division
cycle, and compromising the execution of RB-depen-
dent differentiation programs in certain tissues.

INK4a-ARF: Regulating RB and p53

Figure 1. RB and p53 Regulate Cell Cycle Checkpoint Controls Given the roles of RB and p53 in tumor suppression, it
Mitogenic signals activate cyclin D-dependent kinases, which phos- would be expected that other gene products that act
phorylate RB and RB family proteins (p107 and p130) to facilitate epistatically to regulate their expression or functions
entry into S phase (top). The Cdk2 inhibitor, p27Kip1, expressed at
high levels in quiescent cells, is phosphorylated by cyclin E-Cdk2
might also be frequent targets of deregulation in cancer
in late G1 phase and degraded as cells enter S phase. Constitutive cells. Prominent among these is the INK4 family of Cdk
oncogenic signals can activate the INK4a/ARF locus. By antagoniz- inhibitors, which block the ability of the cyclin D-depen-
ing the activity of cyclin D-dependent kinases, p16INK4a activates RB dent kinases, Cdk4 and Cdk6, to phosphorylate and
and prevents entry into S phase. Mdm2 is a p53-inducible gene that thereby inactivate RB’s growth suppressive functions.
normally acts to terminate the p53 response. The p14ARF protein The founding member, p16INK4a (Serrano et al., 1993) is
inhibits Mdm2 to induce p53, leading either to p53-dependent apo-
ptosis or to induction of the Cdk2 inhibitor p21Cip1, inhibition of cyclin
inactivated in cases of familial melanoma (Kamb et al.,
E/Cdk2, and RB-dependent cell cycle arrest. As cells exit the division 1994) and has since been found to be disabled in many
cycle, p27Kip1 is stabilized and reaccumulates. DNA damage signals tumor types (Ruas and Peters, 1998). Intriguingly, the
activate p53 via ARF-independent pathways. INK4a locus encodes a second, structurally and func-
tionally unrelated protein that is also a potent tumor
suppressor. Two alternative transcripts, initiated at sep-
et al., 1976; Francke and Kung, 1976; Noel et al., 1976). arate promoters and incorporating sequences from dis-
Patients with familial tumors who carried constitutional tinct first exons (designated 1␣ and 1␤), are each spliced
chromosome 13q14 deletions were observed to have a to common downstream exon sequences that are trans-
50% reduction of esterase D activity in their normal cells lated in alternative reading frames. Whereas the tran-
but no remaining activity in their tumor cells, indicating script that contains exon-1␣ sequences specifies p16INK4a,
that esterase D and RB were closely linked. One such the mRNA incorporating exon-1␤ sequences encodes
patient had no detectable deletion of chromosome the alternative reading frame (ARF) protein, designated
13q14 in her normal cells but had a missing chromosome p14ARF in humans and p19Arf in the mouse (Quelle et al.,
13 in her tumor (Benedict et al., 1983). Hence, surrogate 1995). Equally surprising, the ARF protein activates p53
marker analysis had detected a submicroscopic first by binding directly to the p53 negative regulator Mdm2
hit, and subsequent loss of chromosome 13 had likely and protecting p53 from Mdm2-mediated degradation
inactivated the second RB allele. Using restriction frag- (Sharpless and DePinho, 1999; Sherr, 2001) (Figure 1).
ment length polymorphisms, Cavenee et al. (1983) local- Thus, one locus encodes two proteins, p16INK4a and
ized the affected region much more precisely and, im- p14ARF, that functionally interface with RB and p53, re-
portantly, were able to formally conclude that inherited spectively. Despite their intimate linkage, the two genes
and sporadic cases of retinoblastoma lost the same are independently regulated, targeted differentially by
critical allelic sequences. Using probes from the region, various signals, and separately silenced and mutated in
the RB gene was soon cloned (Friend et al., 1986). various forms of cancer. We still do not understand what
We now recognize that RB is part of a gene family purported selective advantage might have led to their
that includes two other members, p107 and p130, which economy of gene organization during evolution, particu-
collectively corepress genes that regulate programs larly since deletions involving INK4a-ARF simultane-
governing cell cycle progression, apoptosis, and differ- ously compromise the functions of both RB and p53.
entiation. Like p53, the RB family proteins exert much The INK4a-ARF promoters respond to sustained hy-
of their growth suppressive control during the G1 phase perproliferative signals. As a singular example, constitu-
of the cell division cycle. RB family proteins physically tive and simultaneous activation of multiple signaling
interact with transcription factors, the best character- pathways by oncogenic Ras induces both Ink4a and
ized of which are the E2Fs. These play key roles in Arf, thereby activating both Rb and p53 and arresting
coordinately regulating many genes required for DNA cell proliferation (Serrano et al., 1997). In contrast, the
Cell
238

loss of Ink4a-Arf extends the replicative capacity of cells the various components undergo phosphorylation by
in culture, contributes to their establishment as continu- glycogen synthase kinase-3␤ (GSK-3␤). Phosphorylated
ously proliferating cell lines, and sensitizes them to ␤-catenin is recognized by an E3 ubiquitin protein ligase
transformation by oncogenic Ras (Serrano et al., 1996; that marks it for degradation by the proteasome (Figure
Kamijo et al., 1997). In short, RB, p53, p16INK4a, and p14ARF 2A, left). Wnt signaling inhibits the enzymatic activity of
form part of a signaling network that monitors mitogenic GSK-3␤, stabilizes ␤-catenin, and enables it to associate
signaling and restrains aberrant growth-promoting sig- with TCF/LEF protein complexes to activate the tran-
nals from driving cell cycle progression inappropriately scription of target genes, including those like c-Myc and
(Figure 1). Inactivation of this signaling network occurs cyclin D1 (CCND1) that promote proliferation (Figure 2A,
in most, if not all, forms of human cancer. right) (Polakis, 1997; Fearnhead et al., 2001). Disruption
of the mouse Tcf7/2 gene, whose product forms tran-
Ligand-Dependent Signaling scriptionally active complexes with ␤-catenin, depletes
and Tumor Suppression intestinal epithelial stem cells, highlighting the role of
A series of genes known to affect positional identity, this signaling pathway in normal intestinal development
tissue patterning, and proliferation during embryonic de- (Korinek et al., 1998).
velopment are also targets of mutations in cancer cells. Mutations that disable APC terminate the polypeptide
Included in this group are genes such as APC, PTCH, chain prematurely, canceling its ability to negatively reg-
SMAD4/DPC4, PTEN, TSC1,2, NF1, and WT1 (Table 1). ulate ␤-catenin turnover and constitutively activating
Many such proteins mediate the flow of information from this signaling pathway (Morin et al., 1997; Korinek et al.,
ligand-dependent cell surface receptors to families of 1997). Mutant forms of ␤-catenin that are resistant to
nuclear transcription factors that govern both develop- phosphorylation and proteasomal turnover have also
mental and proliferative programs. Misregulation of been detected in colorectal cancers (Morin et al., 1997).
these genes might affect a cell’s progression toward a Apart from its role in transcriptional control, ␤-catenin
terminally differentiated, nonproliferating state, thereby associates independently with E-cadherin and aids in
allowing additional mutations to accumulate until a fully cell adhesion. E-cadherin loss has also been observed
tumorigenic phenotype emerges. in epithelial tumors, and its ability to suppress cell trans-
Cell specification within certain tissues is a continu- formation may similarly result from limiting the amount of
ous process that occurs throughout the life of an organ- ␤-catenin available for transcriptional signaling (Gottardi
et al., 2001). Indeed, experimental perturbations that
ism. For example, the processes governing the steady
prevent ␤-catenin from entering the nucleus limit the
formation of blood cells and the rapid renewal of epithe-
proliferation of colon cancer cells (Shih et al., 2000).
lial cells in the intestine and skin rely on functions of
Hedgehog (Hh) proteins (Figure 2B) also play impor-
tissue stem cells that can either self-renew or differenti-
tant roles in patterning decisions, acting from early em-
ate (Weissman, 2000). Although it has been argued that
bryogenesis onward to specify the body plan and organ
between four and seven mutations must occur to trans-
development. In mammals, three Hedgehog ligands
form a normal cell into a tumor cell (Hanahan and Wein-
[Sonic (S), Desert (D), and Indian (I)] are differentially
berg, 2000), the proliferation and terminal differentiation
expressed in various tissues, although Shh is the best
of renewing cell populations in blood, intestine, skin,
characterized and most ubiquitous. Genetic studies of
and other organs, occur on a temporal scale that may
Hh signaling in Drosophila first pinpointed the key sig-
be too rapid to accommodate the multiple mutations
naling components (Murone et al., 1999; Taipale and
required for tumorigenesis. Hence, others have specu-
Beachy, 2001), but their exact biochemical functions
lated that cancers arise from mutations in longer resi-
remain unclear, and mammalian homologs of each of
dent, tissue stem cell populations, thereby shifting the the Drosophila proteins have not been identified. The
balance between self-renewal and differentiation and key upstream target of the signaling pathway in flies is
misspecifying cell fates and numbers within a target the Ci transcription factor, for which three GLI paralogs
organ (Taipale and Beachy, 2001; Reya et al., 2001). have been identified in mammals. In the absence of Hh
Mutations inactivating the APC gene are responsible signaling, Ci/GLI is thought to repress transcription of
for familial adenomatous polyposis (FAP), a disease in target genes, whereas ligand stimulation reverses this
which hundreds of adenomatous polyps arise in the process (Figure 2B). The Hh receptor, Patched, is an
colon and rectum of affected individuals, and where upstream negative regulator of the signaling pathway.
colorectal cancer invariably follows relatively early in the The PTCH gene was identified as a tumor suppressor in
lives of untreated patients. The responsible gene was Gorlin’s syndrome (Gorlin, 1987), where its inactivation is
assigned to chromosome 5q21 and identified by posi- associated with development of basal cell carcinoma
tional cloning (Groden et al., 1991; Kinzler et al., 1991). (BCC) and medulloblastoma (Johnson et al., 1996; Hahn
Although germline mutations in APC account for the et al., 1996a). As for FAP, in which the loss of APC
early appearance of colorectal tumors in FAP patients, predisposes to the appearance of multiple adenoma-
somatic mutations affecting both APC alleles also occur tous polyps, persons inheriting a dysfunctional PTCH
as early events in ⬎80% of sporadic, nonhereditary colo- allele develop numerous BCCs. PTCH mutations have
rectal cancers as well (Powell et al., 1992). also been identified in a significant percentage of pa-
The APC protein interacts with ␤-catenin (Rubinfeld tients with sporadic BCC and medulloblastoma. Again,
et al., 1993; Su et al., 1993), a key component of the the appearance of many differentiated precursor lesions
Wnt/Wingless signaling pathway (Figure 2A). In the ab- in BCC is consistent with the idea that tissue stem cells
sence of a Wnt signal, ␤-catenin associates with a pro- may be targeted.
tein scaffold complex containing APC and axin, in which TGF-␤ signaling (Figure 2C) regulates the expression
Review
239

Figure 2. Tumor Suppressor Proteins Regulating Ligand-Mediated Signaling Pathways

The tumor suppressor proteins APC, PTCH, Smad4/DPC4, PTEN, TSC1,2, and p53 depicted in different images are highlighted by gray shading.
(A) Wnt signaling. In the absence of Wnt ligand (left image), ␤-catenin binds to a destruction complex containing APC, Axin, and GSK-3␤.
Phosphorylation of ␤-catenin facilitates its recognition by a unbiquitin-conjugating E3 ligase (SCFTRCP) that targets it for proteasomal degradation.
When Wnt binds to its receptor (right image), signaling via Frizzled and Disheveled prevents ␤-catenin phosphorylation and destruction. Import
of ␤-catenin and its binding to TCF/LEF transcription factors induces expression of Wnt target genes (adapted from Polakis, 1997; Fearnhead
et al., 2001). Inactivation of APC mimics the effects of the Wnt signal.
(B) Hedgehog signaling. Most of the molecules involved in transmitting the hedgehog (Hh) signal have been genetically identified in Drosophila,
and some as yet have no counterparts in mammalian cells. In the absence of Hh (left image), the Patched receptor (PTCH) negatively regulates
Smoothened (Smo). The Ci transcription factor in flies (homologous to three GLI genes in mammals) is tethered by the kinesin-like molecule,
Costal-2, to a microtubule-anchored cytoplasmic complex. Following its phosphorylation by protein kinase A (PKA), Ci is recognized by a
Slimb-containing ubiquitin E3 ligase and degraded. It is inferred that Gli is subject to similar control. A fragment of Ci is imported into the
nucleus where it binds to DNA to repress target genes. Binding of Hh to Patched relieves suppression of Smo, disrupts the cytoplasmic
complex, and stabilizes Ci/GLI. Hh signaling also antagonizes a suppressor of the Fused kinase (SuFu), leading to phosphorylation (P) of
different components and facilitating Ci/GLI activation (adapted from Murone et al., 1999). Loss of PTCH constitutively activates the pathway.
(C) TGF-␤ signaling. Binding of TGF-␤ to type I (RI) and type II (RII) receptors triggers phosphorylation (P) of RII and its serine-threonine
protein kinase activity. Receptor (R)-Smads 2 and 3 are phosphorylated and relieved from negative regulation by SARA. The R-Smads bind
Smad4 to form a complex that is imported into the nucleus and mediates expression of TGF-␤-responsive genes. Specificity is dictated by
other DNA binding factors (BF) and by either coactivators or corepressors (CO) that determine the nature of the transcriptional response
(adapted from Siegel and Massagué, 2003). Disruption of SMAD4/DPC4 renders cells resistant to inhibition by TGF-␤.
(D) Phosphoinositide 3-kinase (PI3K) signaling. PI3K, which is activated via many growth factor receptors, catalyzes the conversion of
phosphatidylinositol (4,5) bis-phosphate [PI(4,5)P2] to PI (3,4,5)P3. The activity of PI3K is opposed by the PTEN lipid phosphatase. PI(3,4,5)P3
recruits the AKT (PKB) kinase to the plasma membrane where it undergoes phosphorylation by PDK1 (not shown) and activation. AKT
phosphorylates substrates that foster cell cycle progression, cancel apoptosis, and facilitate translation of capped mRNAs. The tuberous
sclerosis complex [TSC1 (hamartin) and TSC2 (tuberin)] antagonizes the function of a G protein (Rheb, not shown) whose activity is required
for activity of the mTOR kinase and its ability to promote translation (adapted from Sulis and Parsons, 2003; Tee et al., 2003). Loss of PTEN
upregulates signaling.

of hundreds of target genes that can coordinately re- apoptosis. The direct phosphorylation of Smad proteins
strain epithelial cell proliferation (Siegel and Massagué, by ligand-activated TGF-␤ receptors facilitates the as-
2003). This endows TGF-␤ with the ability to govern sembly of heterooligomeric Smad transcription factor
complex biological effects, including tissue morphogen- complexes that bind, together with other specificity fac-
esis, angiogenesis, cell migration and adhesion, and tors, to the promoters of TGF-␤-responsive genes. This
Cell
240

can result in downregulation of certain genes, such as mismatch repair gene defect will develop cancer, dis-
c-Myc, that are required for cell proliferation, as well as ease penetrance is high with a lifetime risk of about
the induction of others, such as those encoding Cdk 80%. Errors in DNA replication involve either single base
inhibitors, that slow cell cycle progression (Shi and Mas- mispairing or unfaithful copying of microsatellite DNA
sagué, 2003). Smad-4, a component of the active tran- sequences composed of mononucleotide or dinucleo-
scription complex, was first identified in the guise of a tide repeats. If the latter errors are uncorrected, micro-
tumor suppressor gene deleted in pancreatic cancer satellite instability ensues. The mismatch repair system
(DPC4) (Hahn et al., 1996b). Mutations in SMAD4/DPC4 includes MutS (MSH2, MSH3, MSH6) and MutL (MLH1,
were subsequently identified in colon cancers (Schutte MLH3, PMS1, PMS2) genes involved in mismatch recog-
et al., 1996), and germline mutations are associated with nition and repair, respectively. The MSH2 protein as-
familial juvenile polyposis (Howe et al., 1998). In turn, sembles with either MSH6 or MSH3 to recognize single
mutations affecting the TGF-␤ Type II receptor have also or larger loops of mismatched DNA, respectively, which
been detected in colon cancers (Grady et al., 1999). are then excised by the MLH1/PMS1 complex (Figure
Hence, desensitizing epithelial cells in the pancreas and 3A). Most cases (95%) of HNPCC arise from mutations
colon to the growth inhibitory properties of TGF-␤ in MLH1 and MSH2 (common to both pathways) (Chung
strongly contributes to carcinogenesis. and Rustgi, 2003). Approximately 15% of sporadic colo-
Many receptor systems activate the serine/threonine- rectal tumors also exhibit microsatellite instability most
specific Akt protein kinase (protein kinase B) whose commonly due to epigenetic inactivation of MLH1. Im-
activity enhances protein synthesis, cell growth (mass), portantly, disruption of the mismatch repair system
cell cycle progression, and survival (Figure 2D). Akt acti- leads to a “mutator” phenotype in which resulting ge-
vation depends upon phosphoinositides produced by netic instability ultimately targets other oncogenes and
PI 3-kinase (Cantley, 2002), a process negatively regu- tumor suppressor genes to induce tumor formation
lated by the lipid phosphatase, PTEN (Maehama and (Lengauer et al., 1998; Loeb et al., 2003).
Dixon, 1999; Sulis and Parsons, 2003). PTEN is ubiqui- Like many epithelial tumors, most colon cancers,
tously expressed in eukaryotes, and its inactivation in whether of familial or sporadic origin, exhibit a high
somatic cells results in constitutively elevated levels of degree of aneuploidy resulting from inappropriate seg-
PI(3,4,5)P3. Complete loss of Pten in flies and mice leads regation of chromosomes during mitosis. Normal cells
to early embryonic lethality, and the ability of hypomor- do not progress through mitosis until chromosomes are
phic Akt alleles to rescue the lethality of Pten null Dro- appropriately aligned on the mitotic spindle, but the
sophila embryos emphasizes the importance of Akt as “spindle checkpoint” that monitors the fidelity of this
an effector of this pathway (Stocker et al., 2002). Germ- process is frequently disrupted in those colon cancers
line mutations of PTEN cause four rare human diseases that exhibit chromosomal, as opposed to microsatellite,
with similar clinical features (Sulis and Parsons, 2003), instability (Cahill et al., 1998). Other mitotic miscues or
one of which (Cowden syndrome) is associated with defective cytokinesis can also lead to ploidy changes.
malignant tumor development. Homozygotic inactiva- In the presence of functional p53, such cells arrest in
tion of PTEN occurs frequently in glioblastoma multi- G1 phase, but in its absence, they enter S phase and
forme, endometrial, and advanced prostate cancers. soon become aneuploid after subsequent divisions
Given the pleiotropic effects of this signaling pathway (Lanni and Jacks, 1998; Nigg, 2002). Aneuploidy in ad-
in influencing cell growth (size) and proliferation, as well vanced cancers can also result from telomere dysfunc-
as in countering the effects of other proapoptotic tumor tion, which leads to end-to-end chromosome fusions
and fusion-bridge-breakage cycles that induce p53-
suppressors, such as p53 (Figure 2D), it is not surprising
dependent apoptosis (Artandi et al., 2000). Again, p53
that PTEN inactivation is a frequent event in many forms
loss enables such cells to survive, resulting in the out-
of human cancer.
growth of tumors that exhibit many unbalanced translo-
Cumulatively, these four examples reveal the diversity
cations. Although genes that regulate mitotic progres-
of biochemical mechanisms that can be used to derail
sion, cytokinesis, or telomerase activity in somatic cells
cell growth control signaling pathways. They also illus-
can protect against tumor progression, they have not
trate how the developmental history of particular cell
as yet been implicated in familial cancer syndromes and
types determines the identity of the pathways that are
are not listed in Table 1.
disrupted in different tumor types.
Individuals carrying mutations in a number of other
DNA damage response genes are also highly tumor
The DNA Damage Response prone. Genes regulating such processes that have been
and Genome Instability identified in inherited disorders predisposing to various
Persons carrying germline mutations affecting gene forms of cancer include ATM, NBS1, BRCA1, BRCA2,
products that sense or repair DNA damage are particu- CHK2, and the Fanconi anemia (FA) complex (Figure 3B).
larly prone to cancer. Hereditary nonpolyposis colo- The ATM kinase, whose loss of function results in
rectal cancer (HNPCC or Lynch syndrome) represents ataxia telangiectasia, acts as a sensor of DNA damage,
a family of disorders stemming from mutations in genes being activated specifically in response to double-
required for DNA mismatch repair (Ionov et al., 1993; stranded DNA breaks (Figure 3B). Patients with defec-
Fishel et al., 1993). HNPCC accounts for 1%–3% of all tive ATM function are exquisitely sensitive to ionizing
cases of colorectal cancer, in which patients inheriting radiation, although they show no such sensitivity to cer-
a germline mutation develop colon cancers associated tain other DNA damage signals such as UV irradiation,
with loss of the remaining wild-type allele (Chung and which is sensed by the ATM-related kinase ATR. ATM
Rustgi, 2003). Although not all individuals inheriting a belongs to an evolutionarily conserved family of proteins
Review
241

Figure 3. Tumor Suppressors Involved in the DNA Damage Response

(A) MSH complexes recognize replication errors resulting in single nucleotide (left) or di/trinucleotide (right) mismatched pairing and recruit
MLH complexes that excise mispaired nucleotides, leading to repair (adapted from Chung and Rustgi, 2003).
(B) Double-strand DNA breaks activate the ATM and ATR kinases, which phosphorylate numerous substrates involved in both checkpoint
control and DNA repair. Phosphorylation of p53 and Mdm2 induce G1 phase arrest. NBS1 phosphorylation inhibits late origin firing and
prevents progression through S phase. Phosphorylation of Cdc25C inhibits the activity of cyclin B/Cdk1 to prevent entry into mitosis.
Phosphorylation of NBS1 and BRCA1 trigger their recruitment to DNA damage foci to facilitate homologous and nonhomologous modes of
DNA repair. Activation of BRCA1 also induces p53 (not shown) (adapted from Kastan and Lim, 2000).
(C) DNA crosslinks activate the Fanconi anemia complex (consisting of the Fanc A, C, E, F, and G proteins) which ubiquitylate Fanc D2, enable
its interaction with BRCA2 (equivalent to Fanc D1), and facilitate repair via homologous recombination (adapted from D’Andrea and Grompe,
2003). As indicated, NBS1 enters into complexes with MRE11 and RAD50, whereas BRCA1 binds to BRCA2, which interacts with the RecA
homolog RAD 51.

that also includes the DNA-dependent protein kinase in homologous recombination and nonhomologous end
DNA-PKCS; each of these plays distinct roles in re- joining, respectively, two distinct processes used to re-
sponses to DNA damage. Following DNA double-strand pair DNA breaks (Figure 3B). Inherited BRCA1 and
breakage, ATM phosphorylates a number of proteins BRCA2 mutations are associated with familial breast
(including p53, CHK2 kinase, NBS1, BRCA1, and and ovarian cancers, requiring somatic loss of function
FANCD2) to initiate both cell cycle checkpoint re- of the second allele for tumors to arise. Familial syn-
sponses and DNA repair processes (Figure 3B). Activa- dromes account for 5%–10% of all breast cancer cases
tion of virtually all of the ATM kinase in a cell can be and are typified by early adult onset and a predisposition
induced by very few double-strand DNA breaks, im- to multicentric and bilateral disease. Although most fa-
plying either that ATM is upregulated in response to milial breast cancers are due to mutations of one of the
global damage-induced changes in chromosome struc- two BRCA genes, such tumors can also arise in Li-
ture or through some other amplifying mechanism (Bak- Fraumeni patients and in those with inherited PTEN defi-
kenist and Kastan, 2003). In turn, activation of the ATM ciency (see above). The exact biochemical functions of
substrates p53, CHK2, and NBS1 inhibits cell prolifera- the BRCA proteins remain unclear, although a physical
tion, presumably allowing cells an opportunity to repair association of BRCA1 with BRCA2 and the binding of
damaged DNA. Indeed, CHK2, like p53, is inactivated the latter to RAD51 (a bacterial RecA homolog) at chro-
in some Li-Fraumeni families (Bell et al., 1999), whereas mosomal foci of DNA damage strongly implicate the
NBS1 is mutated in the Nijmegen breakage syndrome, a complex in repairing double-strand breaks by homolo-
disease that closely mimics ataxia telangiectasia (Shiloh gous recombination (Scully and Livingston, 2000; Jasin,
and Kastan, 2001). Although ATR controls a broader 2002). Whether BRCA genes are also required for nonho-
spectrum of DNA damage responses than does ATM, mologous end joining remains controversial.
both ATR and CHK1 are essential genes and, hence, Strikingly, mutations of BRCA genes have not been
not tumor suppressors. observed in sporadic breast or ovarian cancers. If BRCA
Homozygous mutations in ATM and NBS1 predispose loss confers only a weak selective advantage to certain
to lymphomas relatively early in life (Shiloh and Kastan, cell types so that hereditary tumors arise in middle age
2001). Their involvement most likely reflects the require- or later, there may be little opportunity for sporadic ho-
ment for gene rearrangements during early lymphoid mozygous inactivation. Because biallelic BRCA1 loss
development, in which repair errors increase the fre- activates cell cycle checkpoints that trigger proliferative
quency of chromosomal translocations that are the hall- arrest or apoptosis (Scully and Livingston, 2000), any
marks of T and B cell malignancies. ATM is somatically tumors that emerge are likely to have acquired mutations
inactivated in nonfamilial lymphoreticular malignancies, in genes that regulate these processes. Possibly, there
including T cell prolymphocytic leukemia, B-cell chronic may be a restricted temporal and developmental win-
lymphocytic leukemia, and mantle cell lymphoma, fur- dow, during adolescence for example, when epithelial
ther highlighting its role in protecting developing lym- cells in the breast or ovary that sustain such mutations
phocytes from aberrant genomic rearrangements. might not be eliminated (Elledge and Amon, 2002).
The BRCA genes (in a complex with RAD51) and NBS1 Whereas overexpression of cyclin D1 and Her2/Neu,
(in a complex with MRE11 and RAD50) play key functions which occurs frequently in sporadic breast cancers, is
Cell
242

rarely seen in BRCA1-inactivated tumors, p53 mutations

are much more common (Rosen et al., 2003). Such find-
ings suggest a different molecular etiology for the gener-
ation of BRCA1 null and sporadic breast tumors, likely
reflecting the efficacy by which cell cycle checkpoint
controls counteract preexisting genomic instability.
Fanconi anemia complementation groups define eight
different genes, one of which (FANCD1) appears to be
identical to BRCA2 (D’Andrea and Grompe, 2003). This
implicates FA proteins in the process of homologous
recombination, consistent with observations that FA pa-
tients exhibit defective repair of interstrand DNA cross-
links (Figure 3C). FA patients are predisposed to many
types of cancer with acute myelogenous leukemia being
the most common malignancy. Interestingly, ATM-medi-
ated phosphorylation of one of the Fanconi anemia pro-
teins (FANCD2) also plays a role in attenuating DNA
replication by preventing late origin firing (the S phase
checkpoint). Such findings further emphasize the inti-
mate interplay between cell cycle checkpoint control
and DNA repair during the DNA damage response.

Protein Ubiquitination, Turnover, and Suppression

of Tumor Angiogenesis
Tumor growth and metastasis depend upon angiogen-
esis, a process by which quiescent vasculature is in- Figure 4. VHL Regulates HIF Activity to Govern Responses to
duced to sprout new capillaries. Avascular tumors can- Hypoxia
not expand in size because of a lack of blood supply Under normoxic conditions (top) HIF-␣ subunits are hydroxylated on
and oxygen, but their ability to switch on the production both prolyl and asparaginyl residues. Prolyl hydroxylation facilitates
of angiogenic factors explains how they can trigger neo- HIF-␣ recognition by the VHL-containing E3 ligase containing Cul2,
and elongins B and C, whereas asparaginyl hydroxylation prevents
vascularization, maintain oxygen-dependent ATP pro-
binding of p300/CBP coactivators. During hypoxia, unhydroxylated
duction, expand, and metastasize (Hanahan and Folk- HIF-␣ subunits are stabilized, bind to HIF-␤, assemble with coactiva-
man, 1996). tors, and activate transcription (adapted from Kaelin, 2002).
Activation of hypoxia-inducible transcription factor
complexes (HIFs) in response to oxygen deprivation
drives the expression of many genes important for an- Key features exemplified by studies of VHL are that
giogenesis, red cell production, and glycolysis. These most, if not all, of the systems that prevent tumor devel-
include key vascular and hematopoietic growth factors, opment are exquisitely sensitive to levels of protein
such as vascular endothelial growth factor (VEGF), expression, and that much of the relevant circuitry is
erythropoietin, platelet-derived growth factor (PDGF), controlled by ubiquitin-mediated proteolysis. Many path-
and TGF␣ (Semenza, 1999). HIF subunits are transcribed ways governing protein degradation are subverted in
constitutively. However, HIF-␣ proteins undergo prolyl cancer cells. For example, the rates of turnover of
hydroxylation under normoxic conditions, targeting them ␤-catenin and Ci/Gli transcription factors (Figure 2) are
for ubiquitination by the von Hippel-Lindau (VHL) pro- crucial in establishing the thresholds of Shh and Wnt
tein-containing E3 ligase and for subsequent de- required for productive signaling. The p53 negative reg-
struction by the proteasome (Kaelin, 2002). Within this ulator Mdm2 is an E3 ubiquitin ligase that is negatively
multiprotein complex, the VHL protein provides the rec- regulated by Arf binding and ATM phosphorylation. Dis-
ognition motif for prolyl hydroxylated HIF-␣ (Figure 4). ruption of these signaling pathways can be affected by
Under hypoxic conditions, unhydroxylated HIF-␣ sub- mutation of motifs (“degrons”) necesssary for degrada-
units are not recognized by VHL and are therefore stabi- tion (e.g., VHL, ␤-catenin) or by altered expression of
lized, enabling them to form transcriptionally active E3 ubiquitin ligases (e.g., Mdm2 amplification). The fact
complexes with HIF-␤ subunits to drive transcription that changes in protein turnover can either promote or
of hypoxia-induced promoters. Loss of VHL function interfere with a signaling pathway suggests that other
results in constitutive HIF stabilization and predisposes recently discovered components that regulate ubiquiti-
to particular tumors, such as renal clear cell carcinomas, nation will eventually be revealed to act as oncogenes
cerebellar hemangioblastomas, retinal angiomata, and or tumor suppressors.
pheochromocytomas, all of which have a major vascular
component. Tumor formation in the hereditary setting The Basis of Tissue Specificity?
(VHL syndrome) stems from a loss of heterozygosity at Despite our increasing understanding of the cellular
the VHL locus with retention of the mutant VHL allele. functions of many tumor suppressor proteins, we do
Somatic VHL mutations, deletions, and gene silencing not know why the loss of a particular gene contributes
can also be observed in sporadic renal cell carcinomas to tumors in some tissues but not others. Understanding
and cerebellar hemangiosarcomas. the involvement of these proteins in specific biochemi-
Review
243

cal pathways provides some insights. For example, Wnt faster in Kip1 null animals, these data argue that a re-
and Shh signaling are central to developmental pro- duced dosage of p27Kip1, rather than its absolute ab-
grams that affect the formation of the intestine and cere- sence, can contribute to cancer susceptibility. The latter
bellum (Korinek et al., 1998; Taipale and Beachy, 2001; concept has been reinforced in investigations of human
Wechsler-Reya and Scott, 2003), and disruption of these cancers, in which hemizygous loss of Kip1 and/or re-
pathways predisposes to colorectal cancers and medul- duced levels of protein expression confer a poor prog-
loblastoma, respectively. But things are not so tidy. Al- nosis (Blain et al., 2003). There is some evidence that
though inactivation of p16INK4a or RB, which function in other Cdk inhibitors, including p21Cip1, p57Kip2, and p18Ink4c
the same biochemical pathway, is observed in many might function in this manner, so this might prove to be
different tumor types, the fact that small cell lung can- a general property of this class of proteins.
cers preferentially acquire RB mutations while lung ade- Several other examples of haploinsufficiency for tu-
nocarcinomas sustain INK4a loss points to a far greater mor suppression have been demonstrated (Cook and
degree of cell type specificity that defies explanation. McCaw, 2000). For example, mice hemizygous for p53
In the case of the DNA damage response, genes such can develop tumors that retain and express wild-type
as ATM and NBS1 specifically detect double-strand p53 protein (Venkatachalam et al., 1998). In the absence
breaks, and their contribution to lymphoid malignancies of a functional Ink4a allele, animals lacking a single Arf
logically reflects a need for fidelity of repair during the allele are strikingly more prone to melanoma develop-
processes of gene rearrangement that generate the im- ment (Krimpenfort et al., 2001). Pten haploinsufficiency
mune repertoire. But, why then does loss of BRCA1 or in mice accelerates tumor progression, and loss of one
BRCA2, which seem to play central roles in checkpoint PTEN allele with preservation of the second is also com-
and repair responses to DNA damage, specifically pre- mon in human tumors, although its role in these settings
dispose to breast and ovarian cancers? It may prove remains controversial (Sulis and Parsons, 2003). A re-
that a functional redundancy of key signaling pathways
cent study of mice engineered to express a mutant se-
can protect many tissues from dire consequences of
ries of Pten alleles with incrementally decreasing activity
tumor suppressor gene inactivation, whereas particular
and penetrance argues strongly that Pten is haploinsuffi-
cell types that lack such compensatory mechanisms are
cient for tumor suppression in prostate cancer (Trotman
at greater risk to undergo transformation. An improved
et al., 2003). In short, bona fide tumor suppressors such
understanding of tumor suppressor functions, particu-
as p53, Arf, and PTEN may well manifest haploinsuffi-
larly as they relate to processes of tissue-specific ex-
cient effects, particularly when combined with collabo-
pression, cell differentiation, and tissue development
rating mutations affecting additional oncogenes or tu-
will require much more investigation.
mor suppressors.
Haploinsufficiency—When Only One Hit Is Enough
Because of their recessive nature, the traditional ap-
Tumor Susceptibility and Resistance
proach to identifying tumor suppressor genes has been
The vast majority of human cancers show no obvious
to pinpoint small chromosomal regions of loss-of-heter-
familial inheritance, and it has been suggested that mul-
ozygosity (LOH) that occur in particular tumor types
tiple, low penetrance genes segregating in the human
(and frequently in familial cancers), to narrow the critical
population confer cancer susceptibility and resistance
region by deletion mapping, and finally to search the
to environmental carcinogens (Balmain et al., 2003; Loeb
intact homologous chromosomal segment for mutated
et al., 2003). These low penetrance genes might only
genes whose functions can be demonstrated to protect
act combinatorially in a dosage-dependent manner to
against cancer development. However, this strategy will
fail under circumstances in which the second allele is determine cancer predisposition, incrementally affect-
epigenetically silenced or when the targeted gene is ing processes such as carcinogen metabolism, DNA
haploinsufficient for tumor suppression, a situation in repair efficiency, inflammation, and the immune re-
which functional loss of only one allele confers a selec- sponse to provide relative degrees of tumor resistance.
tive advantage for tumor growth (Cook and McCaw, Finding this class of genes presents a great challenge.
2000; Quon and Berns, 2001). Only a few haploinsuffi- In principle, a set of polymorphisms used to define hap-
cient tumor suppressors have been identified so far, lotypes in genes of potential interest can be associated
but this may not reflect their actual number—just the with cancer development, but this approach requires
difficulties in finding them. Indeed, the ease of pin- that the appropriate interacting loci are suspected and
pointing prototypic tumor suppressors within regions of tested. Mouse models of cancer susceptibility provide
LOH seems to have diminished over time, pointing to a way to pinpoint such genes by enriching for combina-
the possibility that a larger haploinsufficient class exists tions of alleles that control a specific disease phenotype,
but still eludes detection (Quon and Berns, 2001; Bal- something that is not feasible in humans. Interbreeding
main et al., 2003). cancer-resistant and -sensitive mouse strains can pro-
One well-defined instance of haploinsufficiency in the vide information about the number and chromosomal
mouse involves the Cdk inhibitor, p27Kip1 (Fero et al., location of genes involved in susceptibility to various
1998). Animals lacking one copy of Kip1 develop tumors forms of cancer (Balmain and Nagase, 1998). Such map-
spontaneously late in life and are highly sensitive to ping is time consuming and labor intensive, but rapid
tumor induction by chemical carcinogens; however, the advances in the human and mouse genome projects
tumors that arise retain the normal Kip1 allele, which have helped to accelerate progress in detecting modi-
encodes a fully functional protein. Although tumors arise fier loci.
Cell
244

In Short... C.J.S. is an investigator of the Howard Hughes Medical Institute

Mechanistic insights stemming from studies of tumor and also acknowledges ALSAC of St. Jude Children’s Research
Hospital for continued support.
suppressor genes, like those gained from work on onco-
genes, have helped to provide us with a conceptual
References
framework for understanding the genetic basis of cancer.
Artandi, S.E., Chang, S., Lee, S.L., Alson, S., Gottlieb, G.J., Chin,
• Prototypic tumor suppressors are recessive. Their L., and DePinho, R.A. (2000). Telomere dysfunction promotes non-
involvement in familial cancer syndromes implies that reciprocal translocations and epithelial cancers in mice. Nature
they are highly penetrant. At the top of the list is p53, 406, 573–574.
being implicated in more than 50% of human cancers. Baker, S.J., Markowitz, S., Fearon, E.R., Willson, J.K.V., and Vo-
In principle, the signaling pathways controlled by gelstein, B. (1990). Suppression of human colorectal carcinoma cell
genes like p53 that are so frequently disrupted in growth by wild-type p53. Science 249, 912–915.
cancer are ideal targets for therapeutic intervention. Bakkenist, C.J., and Kastan, M.B. (2003). DNA damage activates
• Tumor suppressor genes govern a wide range of nor- ATM through intermolecular autophosphorylation and dimer disso-
mal cellular activities. Although their raison d’etre is ciation. Nature 421, 499–506.
not to protect animals against cancer, their involve- Balmain, A., and Nagase, H. (1998). Cancer resistance genes in
mice: models for the study of tumour modifiers. Trends Genet.
ment in cell cycle checkpoint control, mitogenic sig-
14, 139–144.
naling pathways, protein turnover, DNA damage, hyp-
Balmain, A., Gray, J., and Ponder, B. (2003). The genetics and geno-
oxia, and other stress responses illustrates the broad
mics of cancer. Nat. Genet. 33, 238–244.
spectrum of cell-autonomous processes that can be
Bell, D., Varley, J.M., Zydlo, T.E., Kang, D.H., Wahrer, D.C., Shannon,
deregulated in cancer cells. K.E., Lubratovitch, M., Verselis, S.J., Isselbacher, K.J., Fraumeni,
• Despite our appreciation of their biochemical func- J.F., et al. (1999). Heterozygous germ line hCHK2 mutations in Li-
tions, we have no deep understanding of why many Fraumeni syndrome. Science 286, 2528–2531.
tumor suppressors are associated with certain can- Benedict, W.F., Murphree, A.L., Banerjee, A., Spina, C.A., Sparkes,
cers but not others. Possibly, a lack of functional M.C., and Sparkes, R.S. (1983). Patient with 13 chromosome dele-
redundancy of signaling pathways in particular cell tion: evidence that the retinoblastoma gene is a recessive cancer
populations might aggravate the consequences of tu- gene. Science 219, 973–975.
mor suppressor gene inactivation. Moreover, there Bishop, J.M., and Varmus, H.E. (1985). Functions and origins of
may be restricted temporal and spatial windows retroviral transforming genes. In RNA Tumor Viruses (2nd Edition),
R. Weiss, N. Teich, H. Varmus, and J. Coffin, eds. (Cold Spring
throughout organismal development and tissue re-
Harbor, New York: Cold Spring Harbor Laboratory Press), pp.
newal during which mutations that disable these 249–356.
genes can lead to tumor formation. Elucidating the Blain, S.W., Scher, H.I., Cordon-Cardo, C., and Koff, A. (2003). p27
basis of this tissue specificity presents a significant as a target for cancer therapeutics. Cancer Cell 3, 111–115.
future challenge. Cahill, D.P., Lengauer, C., Yu, J., Riggins, G.J., Willson, J.K.V., Mar-
• Some nonclassical tumor suppressors are haploinsuf- kowitz, S.D., Kinzler, K.W., and Vogelstein, B. (1998). Mutations of
ficient. This category likely includes many more genes mitotic checkpoint genes in human cancers. Nature 392, 300–303.
than those so far recognized, but they are much more Cantley, L.C. (2002). The phosphoinositide 3-kinase pathway. Sci-
difficult to identify. ence 296, 1655–1657.
• Lower penetrance genes that determine cancer inci- Cavenee, W.K., Dryja, T.P., Phillips, R.A., Benedict, W.F., Godbout,
dence and response to conventional therapy are likely R., Gallie, B.L., Murphree, A.L., Strong, L.C., and White, R.L. (1983).
to be even more numerous and collectively may be Expression of recessive alleles by chromosomal mechanisms in
retinoblastoma. Nature 305, 779–784.
very important in governing the appearance of spo-
radic cancers in the human population. These poly- Chung, D.C., and Rustgi, A.K. (2003). The hereditary nonpolyposis
colorectal cancer syndrome: genetics and clinical implications. Ann.
genic determinants of cancer would best be called
Intern. Med. 138, 560–570.
tumor susceptibility and resistance genes, since mu-
Comings, D.E. (1973). A general theory of carcinogenesis. Proc. Natl.
tations involving any one of them alone are unlikely Acad. Sci. USA 70, 3324–3328.
to reveal visible phenotypes. In turn, the pathways
Cook, W.D., and McCaw, B.J. (2000). Accommodating haploinsuffi-
that these genes regulate may prove poorer targets cient tumour suppressor genes in Knudson’s model. Oncogene
for therapeutic intervention than those governed by 19, 3434–3438.
more highly penetrant tumor suppressor genes. D’Andrea, A.D., and Grompe, M. (2003). The Fanconi Anaemia/BRCA
pathway. Nat. Rev. Cancer 3, 23–34.
I end where I began. The biochemical role of p53 was
Eliyahu, D., Raz, A., Gruss, P., Givol, D., and Oren, M. (1984). Partici-
discerned less than 15 years ago. There are 30,000 jour- pation of p53 cellular tumour antigen in transformation of normal
nal citations that refer to this one tumor suppressor gene embryonic cells. Nature 312, 646–649.
in the National Library of Medicine. Enough said! Eliyahu, D., Goldfinger, N., Pinhasi-Kimhi, O., Shaulsky, G., Skurnick,
Y., Ara, N., Rotter, V., and Oren, M. (1988). Meth A fibrosarcoma
Acknowledgments cells express two transforming mutant p53 species. Oncogene 3,
313–321.
Writing about tumor suppressor genes was a daunting undertaking
Elledge, S.J., and Amon, A. (2002). The BRCA1 suppressor hypothe-
and was greatly facilitated by information summarized in other ex-
sis: an explanation for the tissue-specific tumor development in
cellent reviews cited in the text and figure legends. I sincerely thank
BRCA1 patients. Cancer Cell 1, 129–132.
James Roberts, Wade Harper, Michael Kastan, John Cleveland,
Martine Roussel, and an anonymous referee for their perceptive Fearnhead, N.S., Britton, M.P., and Bodmer, W.F. (2001). The ABC
criticisms and Joan Massagué for providing an elegant up-to-date of APC. Hum. Mol. Genet. 10, 721–733.
unpublished review about the role of TGF-␤ signaling in cancer. Fero, M.L., Randel, E., Gurley, K.E., Roberts, J.M., and Kemp, C.J.
Review
245

(1998). The murine gene p27Kip1 is haplo-insufficient for tumour sup- Johnson, R.L., Rothman, A.L., Xie, J., Goodrich, L.V., Bare, J.W.,
pression. Nature 396, 177–180. Bonifas, J.M., Quinn, A.G., Myers, R.M., Cox, D.R., Epstein, E.H.,
Finlay, C.A., Hinds, P.W., Tan, T.H., Eliyahu, D., Oren, M., and Levine, and Scott, M.P. (1996). Human homolog of patched, a candidate
A.J. (1988). Activating mutations for transformation by p53 produce gene for basal cell nevus syndrome. Science 272, 1668–1671.
a gene product that forms an hsc70-p53 complex with an altered Kaelin, W.G., Jr. (2002). Molecular basis of the VHL hereditary cancer
half-life. Mol. Cell. Biol. 8, 531–539. syndrome. Nat. Rev. Cancer 2, 673–682.
Finlay, C.A., Hinds, P.W., and Levine, A.J. (1989). The p53 proto- Kamb, A., Gruis, N.A., Weaver-Feldhaus, J., Liu, Q., Harshman, K.,
oncogene can act as a suppressor of transformation. Cell 57, 1083– Tavtigian, S.V., Stockert, E., Day, R.S., III, Johnson, B.E., and Skol-
1093. nick, M.H. (1994). A cell cycle regulator involved in genesis of many
Fishel, R., Lescoe, M.K., Rao, M.R.S., Copeland, N.G., Jenkins, N.A., tumor types. Science 264, 436–440.
Garber, J., Kane, M., and Kolodner, R. (1993). The human mutator Kamijo, T., Zindy, F., Roussel, M.F., Quelle, D.E., Downing, J.R.,
gene homolog MSH2 and its association with hereditary nonpolypo- Ashmun, R.A., Grosveld, G., and Sherr, C.J. (1997). Tumor suppres-
sis colon cancer. Cell 75, 1027–1038. sion at the mouse INK4a locus mediated by the alternative reading
Francke, U., and Kung, F. (1976). Sporadic bilateral retinoblastoma frame product p19ARF. Cell 91, 649–659.
and 13q-deletion. Med. Pediatr. Oncol. 2, 379–385. Kastan, M.B., and Lim, D.-S. (2000). The many substrates and func-
Friend, S.H., Bernards, R., Rogelj, S., Weinberg, R.A., Rapaport, tions of ATM. Nat. Rev. Mol. Cell Biol. 1, 179–186.
J.M., Albert, D.M., and Dryja, T.P. (1986). A human DNA segment Kastan, M.B., Onyekwere, O., Sidransky, D., Vogelstein, B., and
with properties of the gene that predisposes to retinoblastoma and Craig, R.W. (1991). Participation of p53 protein in the cellular re-
osteosarcoma. Nature 323, 643–646. sponse to DNA damage. Cancer Res. 51, 6304–6311.
Gorlin, R.J. (1987). Nevoid basal-cell carcinoma syndrome. Medicine Kern, S.E., Kinzler, K.W., Bruskin, A., Jarosz, D., Friedman, P., Prives,
66, 98–113. C., and Vogelstein, B. (1991). Identification of p53 as a sequence-
Gottardi, C.J., Wong, E., and Gumbiner, B.M. (2001). E-cadherin specific DNA binding protein. Science 252, 1708–1711.
suppresses cellular transformation by inhibiting beta-catenin signal- Kinzler, K.W., Nilbert, M.C., Su, L.K., Vogelstein, B., Bryan, T.M.,
ing in an adhesion-independent manner. J. Cell Biol. 153, 1049– Levy, D.B., Smith, K.J., Preisinger, A.C., Hedge, P., McKechnie, D.,
1060. et al. (1991). Identification of FAP locus genes from chromosome
Grady, W.M., Myeroff, L.L., Swinler, S.E., Rajput, A., Thiagalingam, 5q21. Science 253, 661–665.
S., Lutterbaugh, J.D., Neumann, A., Brattain, M.G., Chang, J., Kim, Knudson, A.G. (1971). Mutation and cancer: statistical study of reti-
S.J., et al. (1999). Mutational inactivation of transforming growth noblastoma. Proc. Natl. Acad. Sci. USA 68, 820–823.
factor ␤ receptor type II in microsatellite stable colon cancers. Can-
Knudson, A.G. (1973). Mutation and human cancer. Adv. Cancer
cer Res. 59, 320–324.
Res. 17, 317–352.
Groden, J., Thliveris, A., Samowitz, W., Carlson, M., Gelbert, L.,
Knudson, A.G., Meadows, A.T., Nichols, W.W., and Hill, R. (1976).
Albertsen, H., Joslyn, G., Stevens, J., Spirio, L., Robertson, M., et
Chromosomal deletion and retinoblastoma. N. Engl. J. Med. 295,
al. (1991). Identification and characterization of the familial adeno-
1120–1123.
matous polyposis coli gene. Cell 66, 589–600.
Korinek, V., Barker, N., Morin, P.J., Van Wichen, D., de Weger, R.,
Hahn, H., Wicking, C., Zaphiropoulous, P.G., Gailani, M.R., Shanley,
Kinzler, K.W., Vogelstein, B., and Clevers, H. (1997). Constitutive
S., Chidambaram, A., Vorechovsky, I., Holmberg, E., Unden, A.B.,
transcriptional activation by a ␤-catenin-Tcf complex in APC⫺/⫺ co-
Gillies, S., et al. (1996a). Mutations of the human homolog of Dro-
lon carcinoma. Science 275, 1784–1787.
sophila patched in the nevoid basal cell carcinoma syndrome. Cell
85, 841–851. Korinek, V., Barker, N., Moerer, P., van Donselaar, E., Huls, G., Pe-
ters, P.J., and Clevers, H. (1998). Depletion of epithelial stem-cell
Hahn, S.A., Schutte, M., Hoque, A.T., Moskaluk, C.A., da Costa, L.T.,
compartments in the small intestine of mice lacking Tcf-4. Nat.
Rozenblum, E., Weinstein, C.L., Fischer, A., Yeo, C.J., Hruban, R.H.,
Genet. 19, 379–383.
and Kern, S.E. (1996b). DPC4, a candidate tumor suppressor gene
at human chromosome 18q21.1. Science 271, 350–353. Krimpenfort, P., Quon, K.C., Mooi, W.J., Loonstra, A., and Berns, A.
(2001). Loss of p16Ink4a confers susceptibility to metastatic melanoma
Hanahan, D., and Folkman, J. (1996). Patterns and emerging ma-
in mice. Nature 413, 83–86.
chanisms of the angiogenic switch during tumorigenesis. Cell 86,
353–364. Lane, D.P., and Crawford, L.V. (1979). T antigen is bound to a host
protein in SV40-transformed cells. Nature 278, 261–263.
Hanahan, D., and Weinberg, R.A. (2000). The hallmarks of cancer.
Cell 100, 57–70. Lanni, J.S., and Jacks, T. (1998). Characterization of the p53-depen-
dent postmitotic checkpoint following spindle disruption. Mol. Cell.
Harbour, J.W., and Dean, D.C. (2000). The Rb/E2F pathway: ex-
Biol. 18, 1055–1064.
panding roles and emerging paradigms. Genes Dev. 14, 2393–2409.
Lengauer, C., Kinzler, K.W., and Vogelstein, B. (1998). Genetic insta-
Harris, H., Miller, O.J., Klein, G., Worst, P., and Tachibana, T. (1969).
bilities in human cancers. Nature 396, 643–649.
Suppression of malignancy by cell fusion. Nature 223, 363–368.
Linzer, D.I., and Levine, A.J. (1979). Characterization of a 54K dalton
Hinds, P.W., Finlay, C.A., Frey, A.B., and Levine, A.J. (1987). Immuno-
cellular SV40 tumor antigen present in SV40-transformed cells and
logical evidence for the association of p53 with a heat shock protein,
uninfected embryonal carcinoma cells. Cell 17, 43–52.
hsc70, in p53-plus-ras-transformed cell lines. Mol. Cell. Biol. 7,
2863–2869. Loeb, L.A., Loeb, K.R., and Anderson, J.P. (2003). Multiple mutations
Hinds, P., Finlay, C., and Levine, A.J. (1989). Mutation is required and cancer. Proc. Natl. Acad. Sci. USA 100, 776–781.
to activate the p53 gene for cooperation with the ras oncogene and Maehama, T., and Dixon, J.E. (1999). PTEN: a tumour suppressor
transformation. J. Virol. 63, 739–746. that functions as a phospholipid phosphatase. Trends Cell Biol.
Howe, J.R., Roth, S., Ringold, J.C., Summers, R.W., Jarvinen, H.J., 9, 125–128.
Sistonen, P., Tomlinson, I.P., Houlston, R.S., Bevan, S., Mitros, F.A., Malkin, D., Li, F.P., Strong, L.C., Fraumeni, J.F., Jr., Nelson, C.E.,
et al. (1998). Mutations in the SMAD4/DPC4 gene in juvenile polypo- Kim, D.H., Kass, J., Gryka, M.A., Bischoff, F.Z., Tainsky, M.A., et al.
sis. Science 280, 1086–1088. (1990). Germ line p53 mutations in a familial syndrome of breast
Ionov, Y., Peinado, M.A., Malkhosyan, S., Shibata, D., and Perucho, cancer, sarcomas, and other neoplasms. Science 250, 1233–1238.
M. (1993). Ubiquitous somatic mutations in simple repeated se- Morin, P.J., Sparks, A.B., Korinek, V., Barker, N., Clevers, H., Vo-
quences reveals a new mechanism for colonic carcinogenesis. Na- gelstein, B., and Kinzler, K.W. (1997). Activation of ␤-catenin-Tcf
ture 363, 558–561. signaling in colon cancer by mutations in ␤-catenin or APC. Science
Jasin, M. (2002). Homologous repair of DNA damage and tumorigen- 275, 1787–1790.
esis: the BRCA connection. Oncogene 21, 8981–8993. Mowat, M., Cheng, A., Kimura, N., Bernstein, A., and Benchimol, S.
Cell
246

(1985). Rearrangements of the cellular p53 gene in erythroleukemic Shih, I.M., Yu, J., He, T.C., Vogelstein, B., and Kinzler, K.W. (2000).
cells transformed by Friend virus. Nature 314, 633–636. The beta-catenin binding domain of adenomatous polyposis coli is
Murone, M., Rosenthal, A., and de Sauvage, F.J. (1999). Hedgehog sufficient for tumor suppression. Cancer Res. 60, 1671–1676.
signal transduction: from flies to vertebrates. Exp. Cell Res. 253, Shiloh, Y., and Kastan, M.B. (2001). ATM: genome stability, neuronal
25–33. development, and cancer cross paths. Adv. Cancer Res. 83,
Nevins, J.R. (2001). The Rb/E2F pathway and cancer. Hum. Mol. 209–254.
Genet. 10, 699–703. Siegel, P.M., and Massagué, J. (2003). TGF-␤ cytostatic and apo-
Nigg, E.A. (2002). Centrosome aberrations: cause or consequence ptotic programs in homeostasis and cancer. Nat. Rev. Cancer 11,
of cancer progression. Nat. Rev. Cancer 2, 815–825. 807–821.
Stanbridge, E.J. (1976). Suppression of malignancy in human cells.
Noel, B., Quack, B., and Rethore, M.O. (1976). Partial deletions and
Nature 260, 17–20.
trisomies of chromosome 13: mapping of bands associated with
particular malformations. Clin. Genet. 9, 593–602. Stehelin, D., Varmus, H.E., Bishop, J.M., and Vogt, P.K. (1976). DNA
related to the transforming gene(s) of avian sarcoma viruses is pres-
Olivier, M., Eeles, R., Hollstein, M., Khan, M.A., Harris, C.C., and
ent in normal avian DNA. Nature 260, 170–173.
Hainaut, P. (2002). The IARC TP53 database: new online mutation
analysis and recommendations to users. Hum. Mutat. 19, 607–614. Stocker, H., Andjelkovic, M., Oldham, S., Laffargue, M., Wymann,
M.P., Hemmings, B.A., and Hafen, E. (2002). Living with lethal PIP3
Oren, M., and Levine, A.J. (1983). Molecular cloning of a cDNA
levels: viability of flies lacking PTEN restored by a PH domain muta-
specific for the murine p53 cellular tumor antigen. Proc. Natl. Acad.
tion in Akt/PKB. Science 295, 2088–2091.
Sci. USA 80, 56–59.
Su, L.K., Vogelstein, B., and Kinzler, K.W. (1993). Association of the
Parada, L.F., Land, H., Weinberg, R.A., Wolf, D., and Rotter, V. (1984).
APC tumor suppressor protein with catenins. Science 262, 1734–
Cooperation between gene encoding p53 tumour antigen and ras
1737.
in cellular transformation. Nature 312, 649–651.
Sulis, M.L., and Parsons, R. (2003). PTEN: from pathology to biology.
Polakis, P. (1997). The adenomatous polyposis coli (APC) tumor
Trends Cell Biol. 13, 478–483.
suppressor. Biochim. Biophys. Acta 1332, F127–F147.
Taipale, J., and Beachy, P.A. (2001). The hedgehog and Wnt signal-
Powell, S.M., Zilz, N., Beazer-Barclay, Y., Bryan, T.M., Hamilton, ling pathways in cancer. Nature 411, 349–354.
S.R., Thibodeau, S.N., Vogelstein, B., and Kinzler, K.W. (1992). APC
Tee, A.R., Manning, B.D., Roux, P.P., Cantley, L.C., and Blenis, J.
mutations occur early during colorectal tumorigenesis. Nature
(2003). Tuberous sclerosis complex gene products, tuberin and ha-
359, 235–237.
martin, control mTOR signaling by acting as a GTPase-activating
Prives, C. (1998). Signaling to p53: breaking the MDM2-p53 circuit. protein complex toward Rheb. Curr. Biol. 13, 1259–1268.
Cell 95, 5–8.
Trimarchi, J.M., and Lees, J.A. (2002). Sibling rivalry in the E2F family.
Quelle, D.E., Zindy, F., Ashmun, R.A., and Sherr, C.J. (1995). Alterna- Nat. Rev. Mol. Cell Biol. 3, 11–20.
tive reading frames of the INK4a tumor suppressor gene encode
Trotman, L.C., Niki, M., Dotan, Z.A., Koutcher, J.A., Di Cristofano,
two unrelated proteins capable of inducing cell cycle arrest. Cell
A., Xiao, A., Khoo, A.S., Roy-Burman, P., Greenberg, N.M., Van Dyke,
83, 993–1000.
T., et al. (2003). Pten dose dictates cancer progression in the pros-
Quon, K.C., and Berns, A. (2001). Haplo-insufficiency? Let me count tate. PLoS Biology 1, 385–396.
the ways. Genes Dev. 15, 2917–2921.
Venkatachalam, S., Shi, Y.P., Jones, S.N., Vogel, H., Bradley, A.,
Reya, T., Morrison, S.J., Clarke, M.F., and Weissman, I.L. (2001). Pinkel, D., and Donehower, L.A. (1998). Retention of wild-type p53
Stem cells, cancer, and cancer stem cells. Nature 414, 105–111. in tumors from p53 heterozygous mice: reduction of p53 dosage
Rosen, E.M., Fan, S., Pestell, R.G., and Goldberg, I.D. (2003). BRCA1 can promote cancer formation. EMBO J. 17, 4657–4667.
gene in breast cancer. J. Cell. Physiol. 196, 19–41. Wechsler-Reya, R.J., and Scott, M.P. (2003). Control of neuronal
Ruas, M., and Peters, G. (1998). The p16INK4a/CDKN2A tumor sup- precursor proliferation in the cerebellum by Sonic hedgehog. Neuron
pressor and its relatives. Biochim. Biophys. Acta.1378, F115–F177. 22, 103–114.
Rubinfeld, B., Souza, B., Albert, I., Muller, O., Chamberlain, S.H., Weinberg, R.A. (1995). The retinoblastoma protein and cell cycle
Masiarz, F.R., Munemitsu, S., and Polakis, P. (1993). Association of control. Cell 81, 323–330.
the APC gene product with beta-catenin. Science 262, 1731–1734. Weissman, I.L. (2000). Stem cells: units of development, units of
Schutte, M., Hruban, R.H., Hedrick, L., Cho, K.R., Nadasdy, G.M., regeneration, and units in evolution. Cell 100, 157–168.
Weinstein, C.L., Bova, G.S., Isaacs, W.B., Cairns, P., Nawroz, H., et
al. (1996). DPC4 gene in various tumor types. Cancer Res. 56, 2527–
2530.
Scully, R., and Livingston, D.M. (2000). In search of the tumour-
suppressor functions of BRCA1 and BRCA2. Nature 408, 429–432.
Semenza, G. (1999). Regulation of mammalian O2 homeostasis by
hypoxia-inducible factor 1. Annu. Rev. Cell Dev. Biol. 15, 551–578.
Serrano, M., Hannon, G.J., and Beach, D. (1993). A new regulatory
motif in cell cycle control causing specific inhibition of cyclin
D/CDK4. Nature 366, 704–707.
Serrano, M., Lee, H.-W., Chin, L., Cordon-Cardo, C., Beach, D., and
DePinho, R.A. (1996). Role of the INK4a locus in tumor suppression
and cell mortality. Cell 85, 27–37.
Serrano, M., Lin, A.W., McCurrach, M.E., Beach, D., and Lowe, S.W.
(1997). Oncogenic ras provokes premature cell senescence associ-
ated with accumulation of p53 and p16INK4a. Cell 88, 593–602.
Sharpless, N.E., and DePinho, R.A. (1999). The INK4A/ARF locus
and its two gene products. Curr. Opin. Genet. Dev. 9, 22–30.
Sherr, C.J. (2001). The INK4a/ARF network in tumour suppression.
Nat. Rev. Mol. Cell Biol. 2, 731–737.
Shi, Y., and Massagué, J. (2003). Mechanisms of TGF-␤ signaling
from cell membrane to the nucleus. Cell 113, 685–700.