THE INK4a/ARF NETWORK IN TUMOUR SUPPRESSION

Transcription

1 THE INK4a/ARF NETWORK IN TUMOUR SUPPRESSION Charles J. Sherr The retinoblastoma protein (RB) and p53 transcription factor are regulated by two distinct proteins that are encoded by the INK4a/ARF locus. Genes encoding these four tumour suppressors are disabled, either in whole or in part, in most human cancers. A complex signalling network that interconnects the activities of RB and p53 monitors oncogenic stimuli to provide a cell-autonomous mode of tumour surveillance. E2F TRANSCRIPTION FACTOR A heterodimeric transcription factor that is composed of an E2F subunit (1 6) and either DP-1 or DP-2. ARF An alternative reading-frame protein of INK4a/ARF.The locus encodes p19 Arf in the mouse and p14 ARF in humans. E3 UBIQUITIN PROTEIN LIGASE The third enzyme in a series the first two are designated E1 and E2 that are responsible for ubiquitylation of target proteins. E3 enzymes provide platforms for binding E2 enzymes and specific substrates, thereby coordinating ubiquitylation of the selected substrates. Department of Tumor Cell Biology, Howard Hughes Medical Institute, St Jude Children s Research Hospital, 332 North Lauderdale, Memphis, Tennessee 38105, USA. sherr@stjude.org The retinoblastoma tumour-suppressor protein (RB) limits cell proliferation by preventing entry into the DNA synthesis (S) phase of the cell-division cycle 1.It does this, at least in part, by blocking E2F TRANSCRIPTION FACTORS (E2Fs) from activating a battery of genes that are needed for DNA replication and nucleotide metabolism 2. Progression into S phase is accelerated when the ability of RB to suppress the E2Fs is disrupted through phosphorylation of RB a process catalysed by cyclin-d- and E-dependent kinases during the first gap phase (G1) of the cell cycle. The p16 INK4a protein directly inhibits the activities of the cyclin D- dependent kinases, CDK4 and CDK6, which maintains RB in its active, anti-proliferative state 3 (FIG. 1a). Functional disruption of the tumour suppressors p16 INK4a or RB or overexpression of the proto-oncogene products cyclin D1 and CDK4 occur in many human cancers, which prompts speculation that disabling the RB pathway is an essential part of the life history of cancer cells 4. The transcription factor p53 is another protein that has an important function in tumour formation. p53 is mutated in more than 50% of human cancers, and accumulates in response to cellular stress from DNA damage, hypoxia and oncogene activation 5,6. When stabilized and activated, p53 initiates a transcriptional programme that can trigger either cell-cycle arrest or suicide 7,8. Genes that are activated by p53 include those that encode the cyclin-e- and A-dependent kinase inhibitor, p21 Cip1, as well as a host of others that are implicated in apoptosis. The p53 protein is also a direct transcriptional activator of its own negative regulator, Mdm2 (HDM2 in humans), which terminates the p53 response through several mechanisms 9. Binding of Mdm2 to p53 antagonizes the transcriptional activity of p53, induces its ubiquitylation, and enforces the export of p53 from the nucleus to the cytoplasm, where it is degraded 10. Mdm2 is, in turn, subject to negative control by the ARF tumour-suppressor protein (p14 ARF in humans and p19 Arf in the mouse). Arf associates directly with Mdm2 to block its ability to interact productively with p53, both by localizing Mdm2 within the nucleolus 11 and by inhibiting Mdm2 s E3 UBIQUITIN PROTEIN LIGASE activity 12. Recent studies indicate that nucleolar sequestration of HDM2 is not essential, but might contribute to, ARFinduced p53 activation 13. HDM2 is overexpressed in 5 10% of human tumours, whereas ARF is silenced or deleted in many others. So, disruption of signalling through the ARF Mdm2 p53 pathway (FIG. 1a) is also a very common feature in cancer. The discovery of ARF was surprising, because its coding sequences partially overlap with those of the INK4a gene (FIG. 1b). Indeed, much of the protein is encoded by an alternative reading frame in INK4a exon 2, from which ARF takes its name 14. Why these two tumoursuppressor genes are so intimately linked chromosomally remains an enigma. However, both are involved in interpreting responses to hyperproliferative signals and in modulating activities of the RB and p53 pathways. The following discussion highlights the roles of INK4a and ARF in tumour surveillance and underscores emerging complexities in the interpretation of their functions and interactions. By convention, uniform capital letters are used to indicate human genes (for NATURE REVIEWS MOLECULAR CELL BIOLOGY VOLUME 2 OCTOBER

2 a b Arf KO Ink4a/Arf 1β 1α 2 3 Ink4a KO Ink4a/Arf KO Ink4a Arf Figure 1 The Ink4a/Arf locus. a The two products of the mouse Ink4a/Arf locus, p16 Ink4a and p19 Arf (p14 ARF in human) indirectly regulate the retinoblastoma protein () and p53, respectively. b Alternative first exons (1α and 1β) that are transcribed from different promoters (arrows) specify the 5 ends of the Ink4a and Arf transcripts, respectively. These are spliced to the same acceptor site in exon 2, which is translated in alternative frames. Ink4a coding sequences in exons 1α, 2 and 3 are denoted by light shading, and Arf coding sequences in exons 1β and 2 are indicated by dark blue shading. The regions that are disrupted in the different knockout (KO) mouse strains are indicated below the figure. The schematic is not drawn to scale, and in both the human and mouse genomes, exons 1α and 1β are separated by >15 kb. (b is adapted from REF. 14.) DP-1 AND DP-2 Proteins that dimerize with E2F subunits, enabling E2F DP complexes to bind to DNA. INK4 FAMILY A family of genes that encode an inhibitor of CDK4. Four such genes designated in order of discovery are INK4a, INK4b, INK4c and INK4d. These encode polypeptides of kda and are designated p16 INK4a, p15 INK4b, p18 INK4c and p19 INK4d, respectively. CIP/KIP FAMILY A family of genes that includes p21 Cip1, p27 Kip1 and p57 Kip2.Cip, CDK-inhibitory protein; Kip, kinase-inhibitory protein. p16 Ink4a Cyclin D1 p19 Arf Mdm2 p53 example, RB) and proteins (RB), and initial capitals to indicate the mouse homologues (such as the gene and protein). FIGS 1, 2 and 4 use the nomenclature from the mouse, as most of the supporting genetic studies have been carried out in this species. Unravelling the RB pathway There is now overwhelming evidence that p16 INK4a acts ubiquitously as a tumour suppressor in human cancers. Soon after the gene was discovered 3, inactivating mutations in INK4a were mapped in familial melanoma families 15, and this spurred several independent investigations that showed INK4a mutations in many types of sporadic cancer 16. The fact that mutually exclusive events in tumours result in the loss of either p16 INK4a or RB, or in the overexpression of cyclin D1 or CDK4,provides the strongest genetic evidence that such a signalling pathway operates in tumour surveillance. The simplest model for this is that RB restricts cellcycle progression by binding to and restraining E2Fs, which, when untethered from RB, promote S-phase entry. However, this is complicated by the fact that three RB-family proteins (the others being p107 and p130) regulate the functions of several E2Fs, each a heterodimer consisting of one E2F subunit (1 6) with a DP-1 or DP-2 binding partner (FIG. 2) 2,17,18. The cyclin D-dependent kinases include six holoenzymes that are assembled from cyclins D1, D2 or D3, and either CDK4 or CDK6. There are also four distinct INK4 FAMILY members that, as well as p16 INK4a, include p15 INK4b, p18 INK4c and p19 INK4d (REF. 19). So far, the known targets of INK4 proteins are the cyclin-d-dependent kinases, the only recognized substrates of which are proteins of the RB family. Although p16 INK4a, cyclin D1, CDK4 and RB are often deregulated in cancer cells, p15 INK4b, p18 INK4c and cyclin D2 have been implicated less often and in fewer types of tumour. Whether p107 and p130, cyclin D3, CDK6 and p19 INK4d are involved in tumour formation is less clear. So, differences in the patterns of expression or activities of these protein family members must account for their differential roles in tumorigenesis. Free E2F DP complexes activate gene expression, but their inhibition by RB-family proteins can either block the activation of E2F or actively repress transcription by recruitment of other negative regulators to E2Fresponsive promoters 2,17,18. E2F6 is an anomaly, having no transactivation domain to which RB-family proteins bind; when associated with a DP subunit, E2F6 binds to DNA, presumably competing with other E2F complexes at promoter sites. Overexpression of E2Fs 1 3 can drive entry into S phase 20,21 by inducing genes that encodes S-phase cyclins and cyclin-dependent kinases (CDKs), DNA polymerase-α, enzymes involved in nucleotide metabolism, and several proteins that function at origins of DNA replication 2,22. Loss of mouse E2f3 alone inhibits activation of many E2f-regulated genes, and greatly slows cell proliferation, whereas loss of E2f1 has no such effects 23. However, both E2f1 and E2f3 can trigger apoptosis 21,24, which is dependent on p53 (see below) and the p53-related protein, p73 (REFS 25,26). RB preferentially interacts with E2Fs 1 3, whereas p107 and p130 associate with E2F4 and E2F5 (FIG. 2).RB normally binds to and therefore inhibits only a small fraction of total E2F activity, and, notably, loss of RB does not completely deregulate the G1 S-phase transition, so p130 and p107 probably compensate. It is widely assumed that the role of RB in tumour suppression depends on the E2Fs, and in the few available mouse models, tumour phenotypes and apoptosis that stem from disruption of -family proteins can be suppressed on an E2f1-null or E2f3-null background 24, Nonetheless, RB associates with many proteins, including other transcription factors that probably enable RB to regulate gene expression in an E2F-independent manner 2,18,30. One of the key arguments that p16 INK4a acts in a linear RB pathway stemmed from observations that p16 INK4a does not effectively inhibit G1-phase progression in cells that lack RB However,we now recognize that the activity of p16 INK4a is tied to that of other CDK inhibitors of the CIP/KIP FAMILY, which oppose the functions of cyclin E CDK2 (REF. 34). Mitogen-dependent formation of cyclin-d-dependent kinases during G1 phase sequesters p27 Kip1 and p21 Cip1, preventing them from inhibiting (FIG. 3). Conversely, INK4 proteins mobilize the bound pool of Cip/Kip proteins from cyclin D CDK complexes, resulting in CDK2 inhibition. Because cyclin E CDK2 cooperates with cyclin D CDK4 with the inactivation of RB, one role of p27 Kip1 is to maintain RB in its active hypophosphorylated state (FIG. 3). CDK2 phosphorylates other substrates that contribute independently to DNA synthesis 34, so it logically follows that expression of the INK4 proteins might affect activities that are unrelated to RB family functions per se. 732 OCTOBER 2001 VOLUME 2

3 Ink4a Ink4b Ink4c Ink4d Figure 2 The Ink4 family. Four Ink4 proteins inhibit six potential cyclin D-dependent kinases that are assembled from cyclins D1, D2 or D3 with either or Cdk6. The three -family members interact preferentially with different E2f subunits, with associating primarily with E2fs 1 3, and p107 and p130 preferentially forming complexes with E2fs 4 and 5. Proteins that are often involved in tumorigenesis are shaded yellow; pink shading denotes proteins which are implicated less frequently; and the green proteins have been implicated only rarely, if at all, in tumour formation. EXTRAMEDULLARY HAEMATOPOIESIS Blood formation at sites outside of the bone marrow, usually in the spleen or liver. EARLY PASSAGE Primary cells explanted into culture, when grown and serially transferred, eventually undergo replicative arrest. Cells at early passage maintain a more robust proliferative capacity, whereas those at late passage replicate less well. PLASMACYTOMA A tumour mass, usually solitary, containing immunoglobulinproducing plasma cells. The presence of many such disseminated tumours, usually in bone, is called multiple myeloma. IMMUNOGLOBULIN PROMOTER ENHANCER Regulatory sequences of genes that encode heavy (µ) or light (κ or λ) immunoglobulin polypeptides that assemble to form antibodies. In B-cell tumours, these regulatory sequences are translocated and fused to other genes, including the Myc oncogene. D1 D2 D3 Cdk6 The INK4 family and cancer It is difficult to understand why INK4a has such a prominent role in tumour suppression as compared with other INK4 family members. In mice, the Ink4a and Ink4b genes, which are closely linked on chromosome 4 (chromosome 9p in humans), are not appreciably expressed during fetal development or in young animals, and it is only as mice age that p16 Ink4a and p15 Ink4b are detected. This raises the possibility that these two genes have been adapted to counter environmentally induced forms of cellular stress that potentially promote tumour formation 19. Recently derived Ink4a-null mice (retaining a functional Arf) are prone to tumour development 35,36. However, in agreement with data indicating that INK4a mutations are much more common than those affecting INK4b in human cancers, Ink4b-null mice neither show overt developmental anomalies nor spontaneously develop many tumours later in life 37.But,disruption ofink4b in the mouse does predispose to EXTRAMEDULLARY HAEMATOPOIESIS and lymphoid hyperplasia 37, consistent with observations that INK4b expression is silenced in certain human myeloid and lymphoid tumours 38. Unlike Ink4a and Ink4b, the Ink4c and Ink4d genes are ubiquitously and abundantly expressed throughout mouse development, and in stereotypic, tissue-specific patterns throughout adult life. Mice lacking functional Ink4d do not develop tumours and enjoy a normal, uncomplicated lifespan. However, male animals lacking both Ink4d and Ink4c have intrinsic defects in germ-cell development and are sterile 39 ; those lacking Ink4d and Kip1 show neuronal dysfunction stemming from an inability of some neurons in the brain to exit the cell cycle during early postnatal development 40. Mice lacking Ink4c spontaneously develop mid-lobe pituitary tumours, which also arise in Kip1-null animals and in heterozygotes. Animals lacking both Ink4c and Kip1 develop mid-lobe pituitary tumours at an accelerated rate, as well as other endocrine neoplasias 41. Therefore, p18 Ink4c, not p16 Ink4a, is the dominant regulator in these cell types. p107 p130 E2Fs 1 3 E2Fs 4,5 Disabling the mouse Ink4a Arf locus Efforts to disable the Ink4a locus were first undertaken before Arf was discovered, and the disruption of Ink4a exons 2 and 3 (FIG. 1b) led to a phenotype that, understandably, was attributed to the loss of p16 Ink4a function 42. The mice spontaneously and rapidly developed cancers of different types in their first year of life. EARLY-PASSAGE mouse embryo fibroblasts (MEFs) from these animals had unusual properties, being able to grow continuously in culture without undergoing senescence, and they were susceptible to transformation by oncogenic Ras without any co-requirement for immortalizing oncogenes, such as Myc or adenovirus E1A. However, when Arf was selectively disabled by disruption of its unique first coding exon (1β), animals that lacked Arf alone and the MEFs that were derived from them showed the cardinal traits of the originally derived Ink4a knockout mice 43. Hindsight indicated that the features initially attributed to loss of p16 Ink4a emanated instead from disabling p19 Arf, begging the question of what a pure Ink4a knockout mouse would be like. This issue has recently been resolved by two groups who used different strategies to disable Ink4a alone, either by knocking in a mutant allele that encodes a defective protein 35 or by generating a partial deletion of exon 1α coding sequences, but leaving cis-regulatory sequences intact 36. Animals that contain the mutant Ink4a allele did not develop any tumours in their first year of life, whereas those with a partial deletion of exon 1α did. The reasons for the discrepancies are uncertain, although it should be noted that the two groups used different embryonic stem-cell lines to target the Ink4a gene. Despite the differences in outcome, both studies showed that the incidence of spontaneous tumours in Ink4a-null animals was considerably less than that seen in Ink4a/Arf-null and pure Arf-null strains. BALB/c mice have a mutated Inka allele that at least partially disables p16 Ink4a function 44. These animals develop PLASMACYTOMAS in response to intraperitoneal injection of pristane oil, indicating that B cells from Ink4a-null mice might be hypersensitive to transformation. Mice that express a Myc transgene driven by the IMMUNOGLOBULIN PROMOTER ENHANCER (Eµ) develop B-cell lymphomas in their first year of life. Placing the Eµ Myc transgene on an Ink4a mutant background did not accelerate B-cell lymphomagenesis 35, whereas lymphoma latency is notably shortened in animals that lack Ink4a/Arf or Arf alone MEFs derived from the pure Ink4a-null strain have no recognizable cell-cycle defects 35,36. Unlike Arf-null MEFs, which proliferate more rapidly than their wildtype counterparts in culture and have properties of permanently established cell lines, the Ink4a-null MEFs have normal cell-cycle profiles, arrest properly on serum withdrawal, and have no colony-forming advantage when plated at low density. Ink4a-null MEFs undergo replicative senescence and cannot be transformed by oncogenic Ras alone 35,36, although in one study, immortalized variants seemed to emerge more quickly 36. Previous experiments in which Ink4a and Arf antisense vectors were introduced into primary MEFs NATURE REVIEWS MOLECULAR CELL BIOLOGY VOLUME 2 OCTOBER

4 p27 E Inactive Active HAPLO-INSUFFICIENT for tumour suppression, in combination with bi-allelic Ink4a loss, as elimination of only one copy of the Arf gene (and continued expression of the other in melanoma cells) seems to programme a different outcome than that observed in Ink4a/Arf-null or Ink4a-null mice. Mitogenic stimulation Ink4a induction D1 p16 D1 p27 p27 Figure 3 Activity of p16 INK4a and CDK inhibitors. Mitogens stimulate the synthesis of cyclin D1 and its assembly to form a complex with, titrating p27 Kip1 (and p21 Cip1, not shown) into complexes with cyclin-d-dependent kinases that are thought to be resistant to Kip1 inhibition. In turn, this frees cyclin-e from negative regulation by p27 Kip1, enabling the kinase to phosphorylate and complete the inactivation of. Induction of p16 Ink4a sequesters into binary complexes, resulting in rapid degradation of cyclin D. The bound pool of p27 Kip1 is released from cyclin-d1 and inhibits cyclin-e, leading to hypophosphorylation and reactivation. Active complexes are shaded yellow., retinoblastoma tumour-suppressor protein. HAPLO-INSUFFICIENCY A state in which loss of only one of two alleles of a gene detectably disables its function. Degraded E Active E Inactive Inactive Active also indicated that Ink4a was insufficient for establishment but facilitated escape from senescence 48. Despite the fact that few of the features now attributable to Arf loss are manifested in pure Ink4a-null mice, p16 Ink4a does act as a tumour suppressor. Ink4a-null animals are particularly susceptible to effects of carcinogens, such as dimethylbenzanthracene (DMBA) and urethane. DMBA-treated Ink4a heterozygotes developed lung tumours, in which expression from the remaining Ink4a allele was silenced 36.Some Ink4a-null mice spontaneously developed melanomas. This is particularly interesting, not only because familial melanoma is associated with p16 INK4a mutations in humans 15, but also because Ink4a/Arf-null and Arf-null strains show little evidence of melanoma development unless crossed with animals that express a Ras transgene under the control of a tyrosinase promoter 49. Strikingly, when an Ink4a/Arf deletion on one chromosome was combined with the mutant Ink4a allele on the other (leaving one wild-type copy of Arf in cis), the incidence of melanomas was greatly increased 35. DMBA treatment resulted in an even higher frequency of highly aggressive melanomas. These data lead to several important conclusions. First, Ink4a loss, although not sufficient to guarantee tumorigenesis, greatly sensitizes mice to carcinogenic insults. Second, melanoma can be successfully modelled in the mouse, with a particularly subtle constellation of founding genetic events that strongly predispose to tumour development. Third, it could prove that Arf is Ink4a and Arf in cellular senescence Like Ink4a and Ink4b, Arf is not noticeably expressed during fetal development. However, explantation of MEFs into culture leads to the rapid induction of these three genes, implying that culture shock per se is all that is required 50,51. Although this probably reflects the nonphysiological milieu of the tissue-culture environment such as incorrect media and growth factors, improper oxygen tension, lack of appropriate matrix or heterotypic cell cell interactions it seems reasonable that these same genes might contribute to stress-dependent cellcycle arrest in vivo 16. MEFs explanted from animals that lack Ink4a or Ink4b alone undergo senescence in culture, but Arf-null MEFs do not. Moreover, the spontaneous establishment of MEF-derived cell lines routinely involves deletion of Arf or mutation of p53 (REF. 43). By contrast, loss of does not bypass senescence, and indeed, all three members of the family must be deleted to ensure MEF immortalization and susceptibility to Ras transformation Therefore, disruption of the Arf Mdm2 p53 pathway is more crucial than subversion of the pathway in bypassing replicative senescence in MEFs. One possible interpretation is that mouse cells rely more on the p53 pathway than on the pathway in their response to culture shock. However, recent data indicate that certain mouse cell types might depend more on Ink4a than Arf loss of function for their establishment as continuously growing cell lines 55. So, a more reasonable view is that Ink4a, Ink4b and Arf contribute to senescence in a cell-type-specific manner. The involvement of INK4a in replicative cellular senescence has been well documented using primary human epithelial cell strains, such as keratinocytes and mammary epithelium 56,57.Recent experiments indicate that culture conditions rather than population doublings per se determine the timing of INK4a induction in passaged human epithelial cells 58. Therefore, there is no evidence to suggest that INK4a functions as part of a telomere-dependent mitotic timer, and the specific signals that are responsible for p16 INK4a induction have yet to be identified 50,51,58. ARF is activated by hyperproliferative signals Arf is induced by oncogenes, such as Myc,adenovirus E1A, mutated Ras and v-abl This results in p53 activation and re-routes cells that have sustained oncogenic damage to alternative fates either growth arrest or apoptosis. Loss of Arf or p53 eliminates this tumoursurveillance mechanism, and instead allows oncogenes to drive proliferation in a manner that favours uncontrolled cell growth and rapid tumour formation. Expression of high levels of Myc in primary MEFs triggers apoptosis, but strongly selects for the emergence of rare immortalized variants that have lost either p19 Arf or p53 function 59. Similarly, in living animals, an intact Arf Mdm2 p OCTOBER 2001 VOLUME 2

5 Atm Ionizing radiation p16 Ink4a Other targets Mitogens Cyclin D1 family E2fs p19 Arf Mdm2 p53 Apoptosis S-phase entry [Twist, Tbx2] p27 Kip1 Cyclin E p21 Cip1 Cell-cycle arrest Figure 4 A signalling network central to tumour suppression. Mitogenic signalling through cyclin-d-dependent complexes helps to cancel suppression, liberating E2fs and stimulating S-phase entry. Cyclin-D-dependent complexes titrate p27 Kip1, enabling cyclin-e to complete phosphorylation (FIG. 3). Because cyclin E is an E2f-responsive gene, the E2f transcriptional programme also reinforces inactivation of. When expressed at abnormally high levels, E2fs induce p19 Arf, inhibiting Mdm2 and initiating a p53 transcriptional response that triggers either apoptosis or cell-cycle arrest. Although opposed by p19 Arf, Mdm2 is a p53-inducible gene that can terminate the p53 response. Activation of p53 inhibits the Arf promoter through as-yet ill-defined mechanisms. Twist and Tbx2 can also repress Arf transcription. Induction of p21 Cip1 by p53 inhibits cyclin-e and returns to its hypophosphorylated, actively inhibitory state. Although p21 Cip1 action connects the Arf Mdm2 p53 axis to -family proteins, p19 Arf can still arrest Cip1-null cells in G1 phase, indicating that p19 Arf also works through other targets. In addition, cells lacking Arf, Mdm2 and p53 are still susceptible to p19 Arf -induced arrest, albeit with prolonged kinetics, again indicating that p19 Arf can interact with proteins other than Mdm2. The Atm kinase responds to ionizing radiation and triggers Mdm2 phosphorylation, as well as both direct and indirect phosphorylation of p53. This abrogates the Mdm2 p53 interaction. In addition, Atm can phosphorylate and activate E2f1., retinoblastoma tumour-suppressor protein. RHABDOMYOSARCOMA Malignant tumour arising from skeletal muscle. pathway protects Eµ Myc transgenic animals from developing B-cell lymphomas Initially, Myc-induced proliferation in B-lymphoid cells is countered by apoptosis. However, lymphomas ultimately arise, most of which sustain Arf deletions or p53 loss. Crossing the Eµ Myc transgene into a heterozygous Arf background accelerates disease, and the functional Arf allele is usually deleted from the tumour cells that emerge 45.Eµ Myc transgenic animals that lack both Arf alleles develop much more flagrant disease and die of highly aggressive lympholeukaemias within only 2 months after birth, underlining Arf s protective role. Compared to Myc, expression of high levels of oncogenic Ras in primary MEFs leads to a different outcome namely, p53-dependent cell-cycle arrest 63. By contrast, Arf-null MEFs are stimulated to proliferate by oncogenic Ras and undergo transformation 43. Although oncogenic Ras induces p19 Arf expression 61,an independent Ras-activated, mitogen-activated protein kinase (MAPK) signalling pathway can induce Mdm2 in Arf-null cells 64. Opposing effects of Ras on p19 Arf and Mdm2 can, therefore, determine the level of p53. This could account for the finding that Ras-transformed cells that retain wild-type p53 are resistant to p53- dependent apoptosis that is induced by DNA damage. Arf is also induced by E2f1, leading to the intriguing idea that p19 Arf provides a connection between the and p53 pathways (FIG. 4) 21,65. Neither E2f1 nor Myc have been demonstrated to bind directly to the Arf promoter. Nonetheless, a prediction is that the effects of mutations within the pathway should be opposed by Arf activation. E2f1, like Myc, is potently apoptotic when overexpressed, and studies with cultured primary MEFs of different genotypes have indicated that the apoptotic functions of Myc might be mediated, at least in part, through the induction of E2f1 (REF 66). However, E2f-1 can also activate p53 and p73 through Arf-independent pathways 25,26, and, in contrast to in vivo studies with Myc (REFS 45 47), we do not yet know whether tumour formation that results from loss of and activation of E2f is accelerated in an Arf-null background. The p53 protein is activated by DNA-damage signals through both p19 Arf -independent and p19 Arf -dependent pathways 6,67. Ionizing radiation triggers the ataxiatelangiectasia, mutated (ATM) kinase-dependent phosphorylation of both p53 and HDM2, which abrogates their functional interaction and triggers a p53 transcriptional response 68,69. Observations that E2f1 is also a substrate of the Atm kinase, and that it is stabilized when phosphorylated in this manner, raise the idea that Atm might also signal to p53 through the E2f1 Arf axis (FIG. 4) 70. This could help to explain how loss of Arf, although not strictly essential for p53 activation, modifies the duration of the DNA-damage response 67. Although an abnormal threshold of mitogenic signals activates Arf, and possibly Ink4a, both are not noticeably expressed in rapidly dividing embryo cells in vivo. The concept that Ink4a/Arf is actively repressed during development in utero stems from work with the Bmi1 repressor, which, when disabled in the mouse germ line, leads to developmental anomalies of the lymphoid and central nervous systems. Strikingly, these defects are reversed on the Inka/Arf-null genetic background 71. Several other gene products, including Twist and Tbx-2, are also potent Arf repressors (FIG. 4) 72,73.The latter genes are overexpressed in human RHABDOMYOSAR- COMAS and breast carcinomas, respectively, arguing that even in the absence of mutations that affect ARF, NATURE REVIEWS MOLECULAR CELL BIOLOGY VOLUME 2 OCTOBER

6 HDM2 or p53, other genetic modifiers can disable the pathway in particular cancers. Signalling loops and branch points Whereas E2f activity might connect the pathway to p53 through Arf, the latter genes might depend on the function of -family proteins to induce cell-cycle arrest (FIG. 4). MEFs lacking, p107 and p130 are resistant to p19 Arf -induced arrest 52,53. The simplest interpretation might be that the ability of p53 to induce the CDK inhibitor, p21 Cip1, enables p19 Arf to prevent phosphorylation of -family proteins. However, p19 Arf can induce G1 growth arrest in primary cells that lack p21 Cip1 (REF. 74), so other proteins must also connect the Arf Mdm2 p53 pathway to. Mutation of p53 or Mdm2 amplification is often accompanied by overexpression of Arf. Conversely, reintroduction of wild-type p53 into such cells suppresses Arf synthesis 75,76. These data provide evidence for feedback control, in which p53 activation dampens Arf function (FIG. 4). The mechanisms for repression, whether direct or indirect, remain undefined, although the implication is that p53, an Arf target, can help to terminate the Arf response, much in the same way that Mdm2, a p53 target, can terminate the p53 response. Finally, the Arf Mdm2 p53 pathway cannot be strictly linear. Mice that lack all three genes develop a spectrum of cancers that are not observed in animals lacking p53 alone, or both p53 and Mdm2 (REF. 77). Moreover, the introduction of p19 Arf into these triplynull MEFs results in cell-cycle arrest in G1 phase, albeit at a much slower rate than in cells containing Mdm2 and p53. Therefore, p19 Arf must have targets other than Mdm2 (FIG. 4). Future directions Many fundamental questions remain unanswered. We do not understand why oncogenes or tumour suppressors that seem to function in the same signalling pathways are differentially targeted in various tumour types. For example, why should RB be eliminated preferentially in small-cell lung cancers, whereas p16 INK4a is disabled in pancreatic carcinomas? Why are certain INK4 or RBfamily members preferentially deregulated in cancer cells whereas others are not? Why does loss of RB in humans predispose to retinoblastoma but yield pituitary tumours in mice; and, in a more general sense, can engineered mice be used to model human tumour development? The key might lie in a better understanding of how particular combinatorial and temporal variations in expression of oncogenes and tumour-suppressor genes promote different tumour types. In this regard, it seems particularly striking that whereas Ink4a/Arf-null mice do not spontaneously develop melanoma 42, animals that lack Ink4a and retain a single Arf allele are highly susceptible 35. Interconnections between the RB and p53 pathways show a complex signalling network that is subject to many inputs and feedback controls (FIG. 4). Genetic experiments indicating that Ink4a and Arf functionally interact in bypassing cellular senescence 48 are consistent with such interactions. We cannot yet interpret how this network responds to yield different and contrasting biological outcomes, such as proliferation on the one hand and cell-cycle arrest or apoptosis on the other. If E2Fs are rate-limiting for S-phase entry but also serve to connect the functions of the RB and p53 pathways, then levels of E2F activity might be important, with a physiological threshold of signals favouring proliferation and an excess of E2F triggering compensatory checkpoint controls. If we assume that the E2f p53 connection is made through p19 Arf, then loss of Arf should interrupt signalling between E2fs and p53 and enhance tumour formation. Importantly, Arf s role in vivo in modulating tumorigenesis in response to loss has not been rigorously tested. We cannot assume on the basis of the available evidence that Myc and E2f will behave in the same manner in this regard. Another emerging idea is that proteins such as Bmi- 1, Tbx2 and Twist actively repress expression of Ink4a/Arf during development, thereby preventing activation of the - and p53-dependent checkpoint mechanisms that might otherwise prohibit rapid cell proliferation in utero. However, observations that Twist and Tbx2 are overexpressed in human tumours imply that inappropriate suppression of ARF in adult tissues increases the probability of tumour formation. Extending this concept, it might well be that all cancers have defects in p53 function, despite the fact that some tumours retain unmutated ARF, HDM2 and p53 genes. On this basis, it is reasonable to wonder whether tumours can ever arise from cells that maintain a truly intact RB and p53 network, and whether the different classes of genetic lesions that affect RB and p53 function in cancer cells predict their responsiveness to genotoxic therapies that are available at present. 1. Weinberg, R. A. The retinoblastoma gene and cell cycle control. Cell 81, (1995). 2. Dyson, N. The regulation of E2F by prb-family proteins. Genes Dev. 12, (1998). 3. Serrano, M., Hannon, G. J. & Beach, D. A new regulatory motif in cell cycle control causing specific inhibition of cyclin D/CDK4. Nature 366, (1993). 4. Sherr, C. J. Cancer cell cycles. Science 274, (1996). 5. Levine, A. J. p53, the cellular gatekeeper for growth and division. Cell 88, (1997). 6. Giaccia, A. J. & Kastan, M. B. The complexity of p53 modulation: emerging patterns from diverging signals. Genes Dev. 12, (1998). 7. Sionov, R. V. & Haupt, Y. The cellular response to p53: the decision between life and death. Oncogene 18, (1999). 8. Bates, S. & Vousden, K. H. Mechanisms of p53-mediated apoptosis. Cell. Mol. Life Sci. 55, (1999). 9. Juven-Gershon, T. & Oren, M. Mdm2: the ups and downs. Mol. Med. 5, (1999). 10. Zhang, Y. & Xiong, Y. Control of p53 ubiquitination and nuclear export by MDM2 and ARF. Cell Growth Differ. 12, (2001). 11. Weber, J. D., Taylor, L. J., Roussel, M. F., Sherr, C. J. & Bar- Sagi, D. Nucleolar Arf sequesters Mdm2 and activates p53. Nature Cell. Biol. 1, (1999). 12. Honda, R. & Yasuda, H. Association of p19 ARF with Mdm2 inhibits ubiquitin ligase activity of Mdm2 for tumor suppressor p53. EMBO J. 18, (1999). 13. Llanos, S., Clark, P. A., Rowe, J. & Peters, G. Stabilization of p53 by p14 ARF without relocation of MDM2 to the nucleolus. Nature Cell Biol. 3, (2001). 14. Quelle, D. E., Zindy, F., Ashmun, R. A. & Sherr, C. J. Alternative reading frames of the INK4a tumor suppressor gene encode two unrelated proteins capable of inducing cell cycle arrest. Cell 83, (1995). 15. Kamb, A. et al. A cell cycle regulator involved in genesis of many tumor types. Science 264, (1994). 16. Ruas, M. & Peters, G. The p16 INK4a /CDKN2A tumor suppressor and its relatives. Biochem. Biophys. Acta Rev. Cancer 1378, F115 F177 (1998). 17. Nevins, J. R. Toward an understanding of the functional complexity of the E2F and retinoblastoma families. Cell Growth Differ. 9, (1998). 736 OCTOBER 2001 VOLUME 2

7 18. Harbour, J. W. & Dean, D. C. The /E2F pathway: expanding roles and emerging paradigms. Genes Dev. 14, (2000). 19. Roussel, M. F. The INK4 family of cell cycle inhibitors in cancer. Oncogene 18, (1999). 20. Lukas, J., Petersen, B. O., Holm, K., Bartek, J. & Helin, K. Deregulated expression of E2F family members induces S-phase entry and overcomes p16 INK4a -mediated growth suppression. Mol. Cell. Biol. 16, (1996). 21. DeGregori, J., Leone, G., Miron, A., Jakoi, L. & Nevins, J. R. Distinct roles for E2F proteins in cell growth control and apoptosis. Proc. Natl Acad. Sci. USA 94, (1997). 22. Muller H. et al. E2Fs regulate the expression of genes involved in differentiation, development, proliferation, and apoptosis. Genes Dev. 15, (2001). 23. Humbert, P. O. et al. E2f3 is critical for normal cellular proliferation. Genes Dev. 14, (2000). Disruption of mouse E2f3, but not E2f1, reduces transcription of E2f-responsive genes and retards entry into S phase. 24. Ziebold, U., Reza, T., Caron, A. & Lees, J. A. E2F3 contributes both to the inappropriate proliferation and to the apoptosis arising in mutant embryos. Genes Dev. 15, (2001). 25. Lissy, N. A., Davis, P. K., Irwin, M., Kaelin, W. G. & Dowdy, S. F. A common E2F-1 and p73 pathway mediates cell death induced by TCR activation. Nature 407, (2000). 26. Irwin, M. et al. Role for the p53 homologue p73 in E2F-1- induced apoptosis. Nature 407, (2000). 27. Yamasaki, L. et al. Loss of E2F-1 reduces tumorigenesis and extends lifespan of 1 +/ mice. Nature Genet. 18, (1998). 28. Pan, H. et al. Key roles for E2F1 in signaling p53- dependent apoptosis and cell division. Mol. Cell 2, (1998). 29. Tsai, K. et al. Mutation of E2f-1 suppresses apoptosis and inappropriate S phase entry and extends survival of deficient mouse embryos. Mol. Cell 2, (1998). 30. Lipinski, M. M. & Jacks, T. The retinoblastoma gene family in differentiation and development. Oncogene 18, (1999). 31. Lukas, J. et al. -dependent cell cycle inhibition by p16 CDKN2A tumor suppressor. Nature 375, (1995). 32. Koh, J., Enders, G. H., Dynlacht, B. D. & Harlow, E. Tumour-derived p16 alleles encoding proteins defective in cell cycle inhibition. Nature 375, (1995). 33. Medema, R. H., Herrera, R. E., Lam, F. & Weinberg, R. A. Growth suppression by p16 Ink4a requires functional retinoblastoma protein. Proc. Natl Acad. Sci. USA 92, (1995). 34. Sherr, C. J. & Roberts, J. M. CDK inhibitors: positive and negative regulators of G 1 -phase progression. Genes Dev. 13, (1999). 35. Krimpenfort, P., Quon, K. C., Mooi, W. J., Loonstra, A. & Berns, A. Loss of Cdkn2a (p16 INK4a ) confers susceptibility to metastatic melanoma in mice. Nature 413, (2001). 36. Sharpless, N. E. et al. Loss of p16 INK4a with retention of p19 ARF predisposes to tumourigenesis in mice. Nature 413, (2001). Specific disruption of Ink4a by point mutation (REF. 35) or by deletion of exon 1 (REF. 36), each with retention of Arf, enables definition of the role of p16 Ink4a in the mouse. Although many features initially attributed to p16 Ink4a instead reflect Arf function, Ink4a does function as a tumour-suppressor gene in the mouse. 37. Latres, E. et al. Limited overlapping roles of p15 INK4b and p18 INK4c cell cycle inhibitors in proliferation and tumorigenesis. EMBO J. 19, (2000). 38. Herman, J. G., Jen, J., Merlo, A. & Baylin, S. B. Hypermethylation-associated inactivation indicates a tumor suppressor role for p15 INK4b. Cancer Res. 56, (1996). 39. Zindy, F. et al. Control of spermatogenesis in mice by the cyclin D-dependent kinase inhibitors p18 Ink4c and p19 Ink4d. Mol. Cell. Biol. 21, (2001). 40. Zindy, F. et al. Postnatal neuronal proliferation in mice lacking Ink4d and Kip1 inhibitors of cyclin-dependent kinases. Proc. Natl Acad. Sci. USA 96, (1999). 41. Franklin, D. S., Godfrey, V. L., O Brien, D. A., Deng, C. & Xiong, Y. Functional collaboration between different cyclindependent kinase inhibitors suppresses tumor growth with distinct tissue specificity. Mol. Cell. Biol. 20, (2000). 42. Serrano, M. et al. Role of the INK4a locus in tumor suppression and cell mortality. Cell 85, (1996). 43. Kamijo, T. et al. Tumor suppression at the mouse INK4a locus mediated by the alternative reading frame product p19 ARF. Cell 91, (1997). 44. Zhang, S., Ramsay, E. S. & Mock, B. A. Cdkn2a, the cyclindependent kinase inhibitor gene encoding p16 INK4a and p19 ARF, is a candidate for the plasmacytoma susceptibility locus, Pctr1. Proc. Natl Acad. Sci. USA 95, (1998). 45. Eischen, C. M., Weber, J. D., Roussel, M. F., Sherr, C. J. & Cleveland, J. L. Disruption of the ARF Mdm2 p53 tumor suppressor pathway in Myc-induced lymphomagenesis. Genes Dev. 13, (1999). 46. Jacobs, J. J. L. et al. Bmi-1 collaborates with c-myc in tumorigenesis by inhibiting c-myc-induced apoptosis via INK4a/ARF. Genes Dev. 13, (1999). 47. Schmitt, C. A., McCurrach, M. E., De Stanchina, E. & Lowe, S. W. INK4a/ARF mutations accelerate lymphomagenesis and promote chemoresistance by disabling p53. Genes Dev. 13, (1999). 48. Carnero, A., Hudson, J. D., Price, C. M. & Beach, D. H. p16ink4a and p19arf act in overlapping pathways in cellular immortalization. Nature Cell Biol. 2, (2000). 49. Chin, L. et al. Cooperative effects of INK4a and ras in melanoma susceptibility in vivo. Genes Dev. 11, (1997). 50. Sherr, C. J. & DePinho, R. A. Cellular senescence: mitotic clock or culture shock. Cell 102, (2001). 51. Wright, W. E. & Shay, J. W. Cellular senescence as a tumorprotection mechanism: the essential role of counting. Curr. Opin. Genet. Dev. 11, (2001). 52. Sage, J. et al. Targeted disruption of the three -related genes leads to loss of G 1 control and immortalization. Genes Dev. 14, (2000). 53. Dannenberg, J.-H., van Rossum, A., Schuijff, L. & te Riele, H. Ablation of the retinoblastoma protein gene family deregulates G 1 control causing immortalization and increased cell turnover under growth-restricting conditions. Genes Dev. 14, (2000). 54. Peeper, D. S., Dannenberg, J.-H., Douma, S., te Riele, H. & Bernards, R. Escape from premature senescence is not sufficient for oncogenic transformation by Ras. Nature Cell Biol. 3, (2001). References 53 and 54 show that whereas MEFs lacking undergo senescence in culture, those lacking several -family members seem to be immortal. Intriguingly, cells lacking, p107 and p130 are resistant to growth arrest by p19 Arf. 55. Randle, D. H., Zindy, F., Sherr, C. J. & Roussel, M. F. Differential effects of p19 Arf and p16 Ink4a loss on senescence of murine bone marrow-derived preb cells and macrophages. Proc. Natl Acad. Sci. USA, 96, (2001). Like MEFs, mouse bone-marrow-derived pre-b cells that lack Arf are immortal, whereas bone-marrowderived macrophages must also silence the Ink4a gene to become established as continuously growing cell lines. 56. Kiyono, T. et al. Both /p16 INK4a inactivation and telomerase activity are required to immortalize human epithelial cells. Nature 396, (1998). 57. Brenner, A. J., Stampfer, M. R. & Aldaz, C. M. Increased p16 expression with first senescence arrest in human mammary epithelial cells and extended growth capacity with p16 inactivation. Oncogene 17, (1998). 58. Ramirez, R. D. et al. Putative telomere-independent mechanisms of replicative aging reflect inadequate growth conditions. Genes Dev. 15, (2001). Growing primary human keratinocytes and mammary epithelial cells over feeder layers in defined medium can prevent induction of p16 Ink4a and enable their immortalization by introducing the telomerase catalytic subunit. 59. Zindy, F. et al. Myc signaling via the ARF tumor suppressor regulates p53-dependent apoptosis and immortalization. Genes Dev. 12, (1998). 60. De Stanchina, E. et al. E1A signaling to p53 involves the p19 ARF tumor suppressor. Genes Dev. 12, (1998). 61. Palmero, I., Pantoja, C. & Serrano, M. p19 ARF links the tumor suppressor p53 to ras. Nature 395, (1998). 62. Radfar, A., Unnikrishnan, I., Lee, H.-W., DePinho, R. A. & Rosenberg, N. p19 Arf induces p53-dependent apoptosis during Abelson virus-mediated pre-b cell transformation. Proc. Natl Acad. Sci. USA 95, (1998). 63. Serrano, M., Lin, A. W., McCurrach, M. E., Beach, D., and Lowe, S. W. Oncogenic ras provokes premature cell senescence associated with accumulation of p53 and p16 INK4a. Cell 88, (1997). 64. Ries, S. et al. Opposing effects of Ras on p53: transcriptional activation of mdm2 and induction of p19 ARF. Cell 103, (2000). Although Ras triggers Arf induction and thereby antagonizes Mdm2, Ras-dependent signalling through an Arf-independent pathway can lead to Mdm2 induction. In the absence of Arf, the latter pathway should dominate, making cells resistant to p53 induction following DNA damage. 65. Bates, S. et al. p14 ARF links the tumour suppressors RB and p53. Nature 395, (1998). 66. Leone, G. et al. Myc requires distinct E2F activities to induce S phase and apoptosis. Mol. Cell 8, (2001). 67. Khan, S. H., Moritsugu, J. & Wahl, G. M. Differential requirement for p19 ARF in the p53-dependent arrest induced by DNA damage, microtubule disruption, and ribonucleotide depletion. Proc. Natl Acad. Sci. USA 97, (2000). DNA damage activates p53 through both Arfindependent and Arf-dependent signalling pathways, so Arf loss affects the durability of the DNA-damage response. 68. Kastan, M. B. & Lim, D. The many substrates and functions of ATM. Nature Rev. Mol. Cell Biol. 1, (2000). 69. Maya, R. et al. ATM-dependent phosphorylation of Mdm2 on serine 395: role of p53 activation by DNA damage. Genes Dev. 15, (2001). 70. Lin, W.-C., Lin, F.-T. & Nevins, J. R. Selective induction of E2F1 in response to DNA damage mediated by ATMdependent phosphorylation. Genes Dev. 15, (2001). The Atm kinase phosphorylates E2f1, enhancing its activity. This indicates a novel mechanism by which ATM might induce p Jacobs, J. J., Kieboom, K., Marino, S., DePinho, R. A. & van Lohuizen, M. The oncogene and polycomb-group gene bmi-1 regulates cell proliferation and senescence through the ink4a locus. Nature 397, (1999). 72. Maestro, R. et al. Twist is a potential oncogene that inhibits apoptosis. Genes Dev. 13, (1999). 73. Jacobs, J. J. L. et al. Senescence bypass screen identified TBX2, which represses cdkn2a (p19arf) and is amplified in a subset of human breast cancers. Nature Genet. 26, (2000). 74. Pantoja, C. & Serrano, M. Murine fibroblasts lacking p21 undergo senescence and are resistant to transformation by oncogenic Ras. Oncogene 18, (1999). 75. Kamijo, T. et al. Functional and physical interactions of the ARF tumor suppressor with p53 and Mdm2. Proc. Natl Acad. Sci. USA 95, (1998). 76. Stott, F. et al. The alternative product from the human CDKN2A locus, p14 ARF, participates in a regulatory feedback loop with p53 and MDM2. EMBO J. 17, (1998). 77. Weber, J. D. et al. p53-independent functions of the p19 ARF tumor suppressor. Genes Dev. 14, (2000). Mice lacking Arf, Mdm2 and p53 show a broader spectrum of tumours than those lacking Mdm2 and p53. Introduction of p19 Arf into primary MEFs that lack Arf, Mdm2 and p53 induces G1 arrest, albeit slowly. Arf must interact with proteins other than Mdm2. Acknowledgments The author gratefully thanks M. F. Roussel and B. Schulman for insightful comments on the manuscript. C. J. S. is an Investigator of Howard Hughes Medical Institute. Online links DATABASES The following terms in this article are linked online to: Locuslink: ATM CDK2 CDK4 CDK6 cyclin D1 cyclin D2 cyclin D3 cyclin E E2F transcription factors HDM2 p15 INK4b p16 INK4a p18 INK4c p19 INK4d p73 INK4a Mouse Genome Informatics: Bmi1 Ink4b Ink4c Ink4d p19 Arf p21 Cip1 p27 Kip1 v-abl Swiss-prot p53 NATURE REVIEWS MOLECULAR CELL BIOLOGY VOLUME 2 OCTOBER