Seminario 7 - 1

REVIEWS

APOPTOSIS AND GENOMIC INSTABILITY Boris Zhivotovsky* and Guido Kroemer ‡ Abstract | Genomic instability is intrinsically linked to significant alterations in apoptosis control. Chromosomal and microsatellite instability can cause the inactivation of pro-apoptotic pathways. In addition, the inhibition of apoptosis itself can be permissive for the survival and ongoing division of cells that have failed to repair DNA double-strand breaks, experience telomere dysfunction or are in an abnormal polyploid state. Furthermore, DNA-repair proteins can regulate apoptosis. So, genomic instability and apoptosis are intimately linked phenomena, with important implications for the pathophysiology of cancer.

GENOMIC INSTABILITY

The failure to transmit an accurate copy of the entire genome from one cell to its two daughter cells. Note that this term does not describe a state but, rather, a process. CENTROSOME

A specialized organelle that duplicates during interphase and that constitutes the centre of the mitotic spindle.

*Institute of Environmental Medicine, Karolinska Institutet, Box 210, Nobels väg 13, SE-171 77 Stockholm, Sweden. ‡ Centre National de la Recherche ScientifiqueUMR8125, Institut Gustave Roussy, 38 rue Camille Desmoulins, F-94805 Villejuif, France. e-mails: [email protected]; [email protected] doi:10.1038/nrm1443

752

Genome integrity and cell proliferation and survival are regulated by an intricate network of pathways that include cell-cycle checkpoints, DNA repair and recombination, and programmed cell death. Permanent or transient GENOMIC INSTABILITY (BOX 1), which represents one of the fundamental characteristics of cancer, might be ascribed to deficiencies in numerous cellular processes including mitotic-checkpoint regulation, and DNA-damage signalling and repair, as well as telomere maintenance and CENTROSOME function. This review will focus on the complex interplay between genomic (in)stability and apoptosis regulation (BOX 2) that participates in carcinogenesis. The relationship between genomic integrity and cell-death regulation can follow at least three different non-exclusive patterns, all of which might be important for the development of cancer. First, genomic instability can lead to the mutation, or altered expression levels, of cell-death regulators. Second, disabled apoptosis can favour genomic instability. Indeed, numerous cellular mechanisms enforce the rule ‘better dead than wrong’, which means that cells that have a damaged genome or are afflicted by many disorders will be aborted by apoptosis — thereby avoiding the propagation of potentially oncogenic mutations. DNA double-strand breaks (DSBs), telomere dysfunction, illicit POLYPLOIDY or abnormal mitoses can directly trigger apoptosis through a default pathway. However, if apoptosis is inhibited for some reason, this increases the risk of

| SEPTEMBER 2004 | VOLUME 5

CHROMOSOMAL INSTABILITY (CIN) at several levels, and cells that are sufficiently fit to survive can be at a growth advantage, which can lead to cancer. Third, a single protein or process might be involved in the control of both apoptosis and genomic instability, which allows ‘crosstalk’ between the processes. Here, we will discuss these different possibilities, their molecular mechanisms and their possible impact on carcinogenesis.

Genomic instability: disabling apoptosis

Genomic instability is a hallmark of cancer, the pathogenesis of which is also characterized by specific genetic and epigenetic changes that can result in defective apoptosis1. It is tempting to assume that genetic instability, after selection, will result in the expansion of a cell population that is relatively resistant to apoptosis induction. Because disabling apoptosis, in itself, might favour genetic instability (see below), it becomes plausible that both mechanisms might cooperate to increase the oncogenic and metastatic potential of transformed cells. During the initial stages of oncogenesis, a series of random alterations in the unstable genome can lead to a collection of nonrandom genetic alterations that affect a restricted set of oncogenes (for example, oncogenes that encode apoptosis inhibitors) and tumour-suppressor genes (which might encode apoptosis facilitators). These genetic alterations would be nonrandom,

www.nature.com/reviews/molcellbio ©2004 Nature Publishing Group

REVIEWS

Box 1 | Distinct types of genetic instability Mutation of caretakers and gatekeepers, enhanced proliferation and telomere attrition

Tetra- or polyploidization due to fusion, endomitosis or failed division

Inappropriate centrosome duplication

End-to-end fusions resulting in dicentric chromosomes

Loss of polyploidy/ tetraploidy checkpoint

Defective spindle checkpoint

Asymmetric division with unequal chromosome distribution

Non-reciprocal translocations due to breakage-fusion-bridge crisis

Selection and re-expression of telomerase

Apoptosis

Selection and adaptation

Structural chromosomal instability

Aneuploid cancer cells

Numeric chromosomal instability

Genetic instability can be subdivided into chromosomal instability (CIN) and microsatellite instability (MSI), the latter being due to mutations in, or silencing of, DNA-mismatch-repair genes (including MLH1, MSH2, PMS1, PMS2 and MLH6). CIN can be classified broadly into numeric CIN (aneuploidy with monosomies, trisomies or higher-order polysomies) and structural CIN (see figure). Structural CIN leads to chromosomal deletions, translocations, homogeneously staining regions (HSRs) or double minutes (acentric chromosomes that lack functional centromeres). Structural instability can result from the defective activation of the DNA-damage checkpoint (the ‘gate keeper’, which involves ATM, ATR, CHK1, CHK2, p53 and so on). In addition, structural CIN is favoured by defective DNA double-strand break (DSB) repair by the mutation of elements of the two repair mechanisms (‘caretakers’). These are non-homologous end joining (NHEJ), which involves DNA-PK, Ku70, Ku80, DNA ligase-4 and XRCC4, or homologous recombination, which involves MRE11/RAD50/NBS1. This results in the generation of broken chromosomes, including the fusion of two centromere-containing chromosomes to form a dicentric chromosome. Such dicentric chromosomes break during mitosis, after which they fuse again, thereby entering the highly unstable breakage-fusion-bridge (BFB) cycle. Similarly, telomere attrition with consequent end-to-end fusions of uncapped chromosomes can lead to the formation of dicentric chromosomes and subsequent BFB cycles. Numeric CIN can result from various events. These include aberrant polyploidization, which occurs, for example, as a result of cell fusion, endomitosis or failed division. The latter occurs, for example, as a consequence of a deficient cytokinesis checkpoint that could be due to mutations in the anaphase-promoting complex. Alternatively, numeric CIN can arise from inappropriate centrosome duplication with multipolar mitosis, which occurs, for example, as a result of cyclin-E overexpression, BRCA1 and BRCA2 mutations, Aurora-A amplification or p53 inactivation. Similarly, it can be caused by defects in the spindle checkpoint, which can occur due to mutations in BUB1, BUBR1 (a homologue of yeast MAD3), MAD1 and MAD2. The main processes that lead to structural CIN or numeric CIN tend to induce apoptosis as a default pathway that aborts cells during the crisis that is associated with replicative senescence — for example, after DNA damage, after polyploidization or after asymmetric cell division. Disabled apoptosis therefore increases the probability that cells which are sufficiently fit to survive can be selected for and generate cancers.

POLYPLOIDY

A situation in which a cell has more than two complete sets of chromosomes in G1 or more than four sets in G2/M: so, triploid cells carry 3N in G1 and tetraploid cells have 4N. Note that polyploidy (not to be confused with hyperploidy, a special case of aneuploidy with too many chromosomes) refers to simple multiples of the normal chromosome number.

because essential and HAPLOINSUFFICIENT genes would have to be maintained in at least one and two copies, respectively. Simultaneously, an increase in the copies of genes that have adverse effects on the metabolism of the cell or on signal transduction would have to be avoided. These constraints apply not only to single genes, but also to adjacent genes and functionally interrelated genes, to maintain LINKAGE DISEQUILIBRIA2. So, the genetic ‘signature’ of a given cancer type can result from the selection of genetic variants of unstable genomes.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

One well-characterized example of how genetic instability can disable apoptosis is provided by the HCT116 colon carcinoma cell line, which is affected by MICROSATELLITE INSTABILITY (MSI) due to mutation of the DNA-mismatch-repair gene MSH2 (mutS homologue-2). MSI often leads to an inactivating frameshift mutation of BAX, a pro-apoptotic member of the BCL2 FAMILY, which results in the increased resistance to apoptosis. This process is involved in tumour progression in vivo, for example when HCT116 cells are injected into immunodeficient mice3, or in vitro, for example after selection for resistance to oxaliplatin (Eloxatin; Sanofi-Synthelabo Inc.)4. Furthermore, inactivating mutations of BAX seem to be under strong selective pressure in colon carcinoma cells with MSI, where the mutation of BAX indicates a poor prognosis for the patient3. CIN can also favour the deletion of master genes that are involved in apoptosis regulation. One example is the deletion of the short arm of chromosome 17 (17p) that contains the p53 gene, and which can lead to the inactivation of p53, provided that the other allele is mutated or deleted. The allelic loss of 17p, in turn, can predispose cells to polyploidy and ANEUPLOIDY in a variety of different cancers5,6, because the inactivation of p53 abolishes a series of checkpoints that maintain genetic stability. These examples point to a complex, dual relationship between apoptosis inhibition and genomic instability. Apoptosis represses CIN after DNA damage

The genome scrambling that is typical of cancer is most likely catalysed by inappropriately repaired DSBs or by eroded telomere ends (see below) that are sensed and processed as DSBs. This is exemplified by the high incidence of cancer in DSB-repair-deficiency syndromes and by the experimental knockout of genes that are involved in DNA repair by homologous recombination and NON-HOMOLOGOUS END JOINING (NHEJ). DSBs that are repaired by the joining of heterologous ends can generate DICENTRIC or ring chromosomes that initiate BREAKAGE-FUSION-BRIDGE (BFB) CYCLES (see below), thereby generating complex non-reciprocal translocations, which are characteristic of human carcinomas. Indeed, such translocations can be oncogenic — because they might carry chimeric or deregulated oncogenes at their breakpoints, or because they can increase the number of copies of an oncogene or delete a tumoursuppressor gene. Under normal conditions, cells that experience a degree of DNA damage that is beyond repair either undergo apoptosis or enter a senescent state (BOX 3). So, when apoptosis is inhibited, cells with altered and potentially unstable genomes, which therefore should have been eliminated, can survive. So, it is important to understand how DSBs can trigger apoptosis. The DNA-damage response is sensed by proteins that belong to the family of phosphatidylinositol 3-kinaselike kinases7,8. This family includes ATM (ataxia telangiectasia mutated), ATR (ATM and RAD3-related) and DNA-dependent protein kinase (DNA-PK). These kinases operate at, or close to, the sites of primary DNA

VOLUME 5 | SEPTEMBER 2004 | 7 5 3 ©2004 Nature Publishing Group

REVIEWS

Box 2 | Morphological and biochemical definition of apoptosis Apoptosis is morphologically defined by cellular and nuclear shrinkage (pyknosis), chromatin condensation, blebbing, nuclear fragmentation (karyorrhexis) and the formation of apoptotic bodies81. At the biochemical level, apoptosis of mammalian cells is characterized by mitochondrial-membrane permeabilization (MMP) and/or massive caspase activation82–84. Two main pathways can lead to apoptosis.

The intrinsic (or stress) pathway MMP is the rate-limiting event in this pathway. MMP causes bioenergetic failure as well as the release of potentially lethal proteins from the mitochondrial intermembrane space. Such lethal proteins include caspase activators (for example, cytochrome c, which activates the APOPTOSOME caspase-activation complex) and caspase-independent death effectors (such as the apoptosis-inducing factor (AIF), which translocates to the nucleus). MMP is regulated, at least in part, by proteins of the BCL2 family 85.

The extrinsic (or death-receptor) pathway This pathway involves the activation of the plasma-membrane receptor of the TNFreceptor superfamily (for example, CD95 (also known as Apo1 or Fas), DR3 and DR4), which leads to the receptor-proximal recruitment of a caspase-activation complex 86. The resulting activation of caspase-8 is either sufficient to trigger the proteolytic activation of other caspases (and to set off a powerful chain reaction) or requires the proteolytic activation of pro-apoptotic proteins of the BCL2 family (in particular BID and BIM), which causes MMP and therefore triggers the ‘mitochondrial amplification loop’. p53 is an upstream inducer of apoptosis that can stimulate the expression of death receptors (for example, DR5 and CD95) and transactivates genes that encode MMPinducing proteins, either from the BCL2 family (for example, BAX, BID, NOXA and PUMA) or other apoptotic regulators (for example, mtCLIC, p53AIP1, ferredoxin reductase, proline oxidase and the BAX adaptor protein ASC)63. So, p53 stimulates the extrinsic and/or the intrinsic pathway, depending on the cellular context. p53 also transrepresses the transcription of MMP inhibitors such as BCL2 and survivin. In addition, p53 can exert transcription-independent pro-MMP effects, either by activating pre-existing BAX (which is pro-apoptotic)21 or by neutralizing BCL2 or BCL-XL (which are anti-apoptotic)20.

CHROMOSOMAL INSTABILITY

(CIN). Genetic changes that are manifested at the level of chromatin maintenance and segregation. HAPLOINSUFFICIENT

A gene that requires bi-allelic expression. Suppression of one allele would reduce the gene dosage below the critical level. LINKAGE DISEQUILIBRIUM

Non-independent assortment of genes during cell division; for example, because they are situated on the same chromosome. MICROSATELLITE INSTABILITY

(MSI). Alterations in the length of short repetitive sequences (microsatellites), which can be detected by the PCR amplification of tumour DNA. After PCR, new bands that were not present in PCR products of the corresponding normal DNA will appear if MSI has occurred.

754

damage: ATM responds preferentially to DSBs, and ATR to single-stranded breaks that are coated with replication protein A. In human cells, the interaction of RAD17 protein with the 9-1-1 complex (which comprises RAD9, RAD1 and HUS1) is responsible for the recognition of damaged DNA9. This results in the activating phosphorylation of downstream kinases (such as the checkpoint kinases (CHK)1 and 2) by ATM/ATR10. CHK2, in turn, can phosphorylate p53, thereby increasing the DNA-binding activity of p53 as well as its stability. The relative contribution of these various phosphorylation events to physiological DNAdamage-induced apoptosis versus DNA repair have been difficult to assess. But experiments have shown that the expression of a dominant-negative mutant form of CHK2 inhibited p53-mediated apoptosis11. Moreover, CHK2-deficient mice show significantly reduced radiation-induced apoptosis in neurons, thymocytes and splenocytes12, which indicates that CHK2 mediates cell-cycle arrest, thereby facilitating DNA repair while inhibiting apoptosis. Another target for ATM and ATR is H2AX, a minor histone-H2A variant, which undergoes specific rapid phosphorylation on Ser139 following DNA damage. Phosphorylation of H2AX at DSBs is essential for the recruitment of several proteins that are involved in the regulation and/or the execution of DNA repair,

| SEPTEMBER 2004 | VOLUME 5

namely 53BP1 (p53-binding protein-1), BRCA1 (breast-cancer-associated protein-1), MDC1 (mediator of DNA-damage checkpoint-1) and the MRE11 (meiotic recombination-11)–RAD50–NBS1 (Nijmegen breakage syndrome-1) complex to the site of DNA damage10. H2AX-deficient mouse embryonic fibroblasts manifest defects in the localization of DNA-repair factors to stable foci at DSBs and develop CIN after DNA damage13. Conversely, the phosphorylation of H2AX is an acute consequence of DNA damage and a predictor of DNA-damage-induced cell death14. Surprisingly, DSBs that are induced by γ-irradiation do not only cause acute effects but can also lead to delayed cell death, as late as 30–35 generations after the initial DNA lesion. This radiation-induced delayed cell death is accompanied by the appearance of foci of the phosphorylated histone H2AX (which indicates the presence of DNA lesions), as well as by the activating phosphorylation of CHK2 and p53 and the consequent reactivation of p53-inducible genes15. It is still a matter of speculation whether genomic instability that results, for example, from BFB cycles can account for this delayed cell death. The tumour-suppressor protein p53 mediates part of the response of mammalian cells to DNA damage, either by stimulating DNA repair or — beyond a certain threshold of DNA damage — by initiating apoptosis (FIG. 1). p53, which is a transcription factor, transactivates a series of pro-apoptotic proteins from the BCL2 family (in particular BAX, BID, PUMA and NOXA)16, which induce MITOCHONDRIAL-MEMBRANE PERMEABILIZATION (MMP) and therefore release apoptogenic factors from the mitochondrial intermembrane space. p53 upregulates the adaptor protein ASK (activator of S-phase kinase; which promotes the activation of BAX and its interaction with mitochondria), as well as several proteins that locate to mitochondria. These proteins favour MMP through oxidative reactions, such as those that are catalysed by ferredoxin reductase and proline oxidase, or through unknown mechanisms, such as those that are mediated by p53AIP1 (p53-regulated apoptosisinducing protein-1) and mtCLIC (mitochondrial chloride intracellular channel-4). p53 also represses the anti-apoptotic protein BCL2, which works on mitochondria to prevent membrane permeabilization. In addition, p53 can initiate apoptosis through proteins that localize to the endoplasmic reticulum (ER; for example, Scotin) or the plasma membrane (such as CD95, DR5 and PERP). Finally, in response to DSBs, p53 can somehow stimulate the nuclear release of histone H1.2, which then works on mitochondria to stimulate MMP17. So, p53 can engage several, in part celltype-specific, pro-apoptotic pathways and promotes cell death by transactivating a wide array of apoptosisinducing genes16. p53 can also induce apoptosis in a transcription-independent manner18, although to what extent this is important for DNA-damage-induced apoptosis is controversial. p53 has been reported to bind to the outer mitochondrial membrane and antagonize the anti-apoptotic function of BCL2 and BCL-XL19,20. Importantly, some p53 mutants www.nature.com/reviews/molcellbio

©2004 Nature Publishing Group

REVIEWS

Box 3 | Senescence and apoptosis In the 1960s, L. Hayflick observed that non-transformed cells undergo only a limited number of passages, and then enter a state that he called ‘cellular senescence’87. Although proliferation of these cells was arrested, senescent cells were metabolically active and could be maintained in culture for several years. The so-called ‘Hayflick limit’ was linked to the progressive shortening of chromosomal ends during cell division88. The senescence phenotype is associated with the upregulation of several marker proteins, such as INK4A, ARF, p53, PML (promyelocytic leukaemia), clusterin and plasminogen-activator inhibitor. Biochemical and morphological features of replicative senescence (that is, the situation when telomere shortening has reached a critical length) are similar to those observed during ‘premature’ cellular senescence, which is induced by DNA-damaging agents and other types of anti-cancer therapy. Senescence of fibroblasts is associated with resistance to radiation-induced apoptosis89. Moreover, some senescent fibroblasts are unable to undergo p53-dependent apoptosis and, on DNA damage, become necrotic90. However, resistance to apoptosis is not a general feature of senescent cells, as they can also be apoptosis prone: it is cell-type and stimuli dependent. Senescence has been proposed as a mechanism to block immortalization and tumorigenesis. Moreover, in addition to apoptosis, premature or inducible senescence was identified as an effective response to chemotherapy91. Interestingly, the inactivation of ARF, INK4a or p53 is sufficient to disable both senescence and apoptosis. Consistently, activation of the ARF–p53 pathway is important for the effective removal of cancer cells. Recently, it was proposed that the simultaneous stimulation of a mitogen-activated pathway and the downstream inhibition of cyclin-dependent kinases might lead to cell senescence92. This model can distinguish between two types of growth arrest: first, exit to G0 phase, which can be the result of mitogen withdrawal and might lead to apoptosis; and second, so-called hypermitogenic arrest, which might be stimulated by mitogens and lead to senescence. Additional research is required to test this hypothesis.

BCL2 FAMILY

A family of proteins that all contain at least one BCL2 homology (BH) region. The family is divided into antiapoptotic multidomain proteins (such as BCL2 and BCL-XL), which contain four BH domains (BH1, BH2, BH3, BH4), pro-apoptotic multidomain proteins (for example, BAX and BAK), which contain BH1, BH2 and BH3, and the pro-apoptotic BH3-only protein family (for example, BID, BIM and PUMA). ANEUPLOIDY

The ploidy of a cell refers to the number of chromosome sets that it contains. Aneuploid karyotypes are chromosome complements that are not a simple multiple of the haploid set. NON-HOMOLOGOUS END JOINING

(NHEJ). The main pathway that is used throughout the cell cycle to repair chromosomal doublestrand DNA breaks in somatic cells. In contrast to homologousrecombination repair, NHEJ is error-prone because it leads to the joining of heterologous ends.

simultaneously lose the capacity to bind to DNA and to BCL2 and BCL-XL (REF. 20). Alternatively, p53 might activate the pore-forming, MMP-inducing function of BAX21 or BAK22. So, p53 triggers numerous transcriptiondependent and transcription-independent pathways that link DNA damage to apoptosis. This seems important in light of the fact that the genetic or functional inactivation of p53 can lead to genomic instability (see below). The pro-apoptotic and cell-cycle-arresting functions of p53 have been attributed to distinct transcriptional profiles (for example, increased transcription of BAX for apoptosis induction versus increased transcription of p21 for cell-cycle arrest). These profiles correlate, to some extent, with the phosphorylation of p53 on Ser46 (which augments its pro-apoptotic potential)16. Whether such phosphorylation events also affect the non-transcriptional effects of p53 is currently unknown. There are growing suspicions that p53-independent events might link DNA damage to apoptosis induction. Such links could involve p73, which functions as a p53-related transcription factor23, and NUR77 (also known as NGF1β or TR3), which is an orphan steroid receptor that can translocate to mitochondria and specifically interact with BCL2, thereby inducing MMP24. There might also be a link through CASPASE-2 — which can be activated in the nucleus by the ‘PIDDosome’, a molecular complex that contains the protein PIDD (a p53-inducible, deathdomain-containing protein) and the protein RAIDD/CRAIDD (an adaptor protein containing a death domain). Activated caspase-2 can then trigger the mitochondrial apoptotic pathway25. However, it is

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

unclear at present to what extent, if at all, the inhibition of p73, NUR77 or caspase-2 might contribute to genomic instability. Telomeric instability and apoptosis

Telomeres cap the ends of eukaryotic chromosomes, thereby preventing them from being recognized as DSBs that can be processed by DNA-repair mechanisms. They comprise tandem repeats of the TTAGGG sequence and the T-loop (which is formed by the invasion of the single-stranded 3′ overhang into an upstream doublestranded region of the telomere), as well as an array of protective telomere-binding proteins (including telomeric-repeat-binding factor (TRF)1, TRF2, Ku86 and the DNA-PK catalytic subunit (DNA-PKcs)). The loss of TTAGGG repeats increases with ongoing cell division, unless the telomerase (which comprises an RNA molecule known as TERC and a catalytic protein subunit known as TERT) is activated. Telomere attrition or mutation of telomere-binding proteins causes telomere uncapping with the subsequent activation of the telomere checkpoint responses (that is, replicative SENESCENCE and apoptosis) or, alternatively, chromosome end-to-end fusions, which results in karyotypic disarray. Mechanistically, telomere attrition ultimately prevents the formation of the T-loop, which leads to the formation of an unprotected chromosome end that is detected as damaged DNA and processed by the DNA-repair machinery (in particular the NHEJ machinery)26. So, telomere dysfunction and the associated formation of dicentric chromosomes can set BFB cycles in motion, which produce extensive and rapid changes in gene dosage as well as complex cytogenetic alterations (structural CIN)27. Moreover, telomere dysfunction can cause cytokinesis to fail, which results in tetraploidization and the accumulation of supernumerary centrosomes, thereby causing numeric CIN (BOX 1)28. Inactivation of the telomere checkpoint, coupled with growth beyond the Hayflick limit (the lifespan of normal fibroblasts in vitro; see BOX 3), leads to a ‘crisis’ — a period of severe telomere dysfunction that is accompanied by rampant genomic instability and massive cell death 29 (FIG. 2). The absence of the retinoblastoma (Rb)/INK4a/p53-dependent senescence checkpoint (BOX 3) prevents some of the phenomena that are elicited by short or dysfunctional telomeres, such as apoptosis or cell-cycle arrest, but it has no effect on end-to-end fusions in vitro. In vivo, the knockout of the p53 gene (and to some extent also that of the p53-activating kinase that is encoded by the Atm gene) reduces apoptosis in the gastrointestinal crypts of mice that lack the telomerase RNA component Terc. But this does not suppress the generation of dicentric chromosomes30. So, p53 inactivation enhances the survival of cells with short, dysfunctional telomeres and therefore generates wholesale genomic changes that might drive the neoplastic process31. Importantly, it appears that only after reactivation of the telomerase, which quells severe CIN, does a subset of post-crisis cells emerge with a genetic profile that is permissive for malignant progression. So, the role of telomerase in carcinogenesis is complex, in that the

VOLUME 5 | SEPTEMBER 2004 | 7 5 5 ©2004 Nature Publishing Group

REVIEWS

p53-dependent

p53-independent

DNA damage

Caspase-2 Mitochondrion

P

BAX H1.2

DICENTRIC CHROMOSOME

A chromosome that carries two centromeres, which arise from the aberrant fusion of ‘naked’ telomeres or interstitial doublestrand breaks.

H2AX RAD17

NOXA

ATM/ATR

NUR77 BCL-XL

Dicentric chromosomes — which are generated by the fusion of two centrosomecontaining chromosome fragments or two ‘naked’ telomeres — break apart in mitosis, after which they fuse again, forming a mutated dicentric chromosome. APOPTOSOME

A complex that forms when cytochrome c is released from mitochondria and interacts with the cytosolic protein APAF1, which, in turn, recruits procaspase-9. In the presence of dATP, this interaction results in the allosteric activation of caspase-9 and in the formation of a caspase-3 activation complex. MITOCHONDRIAL-MEMBRANE PERMEABILIZATION

(MMP). A pro-apoptotic process whereby mitochondrial membranes undergo transient and stable permeabilization and become structurally reorganized. As a result, proteins that are normally retained in the intermembrane space are released through the outer membrane and the bioenergetic function of mitochondria is compromised. CASPASE

A family of cysteine proteases that cleave after asparagine residues. Initiator caspases are typically activated in response to particular stimuli (for example, caspase-8 after death-receptor ligation, caspase-9 after apoptosome activation, caspase-2 after DNA damage and PIDD activation), whereas effector caspases (mainly caspase-3, -6 and -7) are particularly important for the ordered dismantling of vital cellular structures.

756

?

9-1-1

PUMA BCL2

BREAKAGE-FUSION-BRIDGE (BFB) CYCLE

BAX

p73 PUMA

BAX BAK

CHK2

p53

P

Cytochrome c

Cytochrome c Nucleus Cytoplasm

Pro-caspases

Figure 1 | p53-dependent and p53-independent processes that link double-strand DNA breaks to the apoptotic default pathway. p53-mediated apoptosis might require the transcriptional activation of several genes that function at the level of the plasma membrane (including CD95; not shown in the figure) or the mitochondria (NOXA or PUMA) to induce apoptosis. Alternatively, extranuclear p53 protein can function in a transcription-independent manner, by physical interaction with multidomain proteins from the BCL2 family, which leads to the neutralization of anti-apoptotic proteins such as BCL2 or BCL-XL and/or to the activation of pro-apoptotic proteins BAX and BAK. These effects are, in part, mediated by transcription-independent mechanisms and direct physical interactions between p53 and members of the BCL2 and the BAX family (dashed arrows). In addition, damaged nuclei can release histone H1.2 in a p53-dependent fashion, and histone H1.2 then works on mitochondria. Alternatively, DNA damage can trigger apoptosis by p53-independent routes that might involve caspase-2, NUR77 and p73. Whether the pathway that links DNA damage to apoptosis is p53-dependent or p53-independent, the activation of the mitochondria-mediated pathway is essential for programmed cell death. In the centre of the diagram, a series of proteins that are involved in the formation of DNA-damage foci (ATM (ataxia telangiectasia mutated)/ATR (ATM- and Rad3-related), RAD17, histone H2AX, 9-1-1 and CHK2 (checkpoint kinase-2)) are linked with both DNA-repair and apoptotic machineries.

absence of telomerase activity must be followed by the reacquisition of telomerase function to guarantee effective immortalization after genome scrambling27. Importantly, it is possible that the telomerase itself has an anti-apoptotic function, which works at the pre-mitochondrial level32. This anti-apoptotic function operates in vivo, as transgene-enforced expression of the catalytic subunit of telomerase protects against brain injury that is caused by ischaemia33. In theory, pharmacological inhibition of telomerase could therefore have a direct apoptosis-facilitating effect on cancer cells. Ploidy control and apoptosis

Polyploidy can be generated by cell fusion (BOX 4) or, in a cell-autonomous fashion, by the multiplication of chromosomes without accompanying cellular division34. This could involve ENDOREPLICATION, as seen in megakaryocytes, which undergo successive rounds of DNA replication accompanied by incomplete mitoses (which proceed through anaphase A but omit anaphase B and cytokinesis), resulting in mononuclear polyploidy35. Alternatively, it can involve an abortive cell cycle in which, for example, nuclear division is not

| SEPTEMBER 2004 | VOLUME 5

accompanied by cytokinesis, which produces binucleate cells36. In addition, tetraploidization can be induced by microtubule-stabilizing agents such as taxol or nocodazole37. Following spindle damage, cells become transiently arrested at the metaphase–anaphase transition point and then escape from the block, a process known as ‘mitotic slippage’, and exit mitosis without proper segregation of sister chromatids and cytokinesis (FIG. 3). Cells that have a damaged spindle then arrest permanently in a tetraploid G1 state. This final arrest is mediated through p53, whereas the transient arrest at the metaphase–anaphase transition and mitotic slippage are probably not influenced by p53. p53-deficient cells that are in the tetraploid G1 state are not prevented from re-entering the cell cycle to reduplicate their DNA unchecked, which leads to polyploidy and subsequent CIN. In the presence of p53, tetraploidy (or polyploidy in general) causes the activation of p21 and an irreversible arrest in the cell cycle, or causes cell death, thereby preventing the propagation of errors of late mitosis and the generation of aneuploidy38,39. The absence of p21 also relaxes the polyploidy checkpoint and causes

www.nature.com/reviews/molcellbio ©2004 Nature Publishing Group

REVIEWS

Cell death

p53

Senescence and apoptosis

End-to-end joining, BFB cycles

Telomere attrition or dysfunction Genomic instability

Re-expression of telomerase Cancer

Figure 2 | Telomeric instability. When telomeres become dysfunctional, ‘naked’ telomeres are recognized as double-strand DNA breaks and trigger the p53-dependent apoptotic default pathway. In the absence of functional p53, genomic instability results from futile end-to-end joining and breakage-fusion-bridge (BFB) cycles, which normally result in cell death. Only upon re-expression of telomerase can immortal cancer cells be formed.

SENESCENCE

A nearly irreversible stage of permanent G1 cell-cycle arrest, which is linked to morphological changes (flattening of the cells), metabolic changes and changes in gene expression (for example, β-galactosidase). The induction of senescence depends on p53 and cell-cycle inhibitors such as p21 and p16. ENDOREPLICATION

The replication of DNA during S phase of the cell cycle without the subsequent completion of mitosis. PASSENGER PROTEIN

A protein that shares a characteristic pattern of association with chromatin in prophase, centromeres in metaphase and early anaphase, and then the midzone and midbody in late anaphase and telophase, respectively. ANISOCYTOSIS

Abnormal heterogeneity in cell size. ANISOKARYOSIS

Abnormal heterogeneity in nuclear size and/or in the cytoplasm:nucleus ratio. DNA-STRUCTURE CHECKPOINT

A checkpoint that arrests cellcycle advancement until DNA mutations such as double-strand breaks are repaired, or until the replication of complementary strands has been completed.

haematopoietic MO7e cells to manifest centrosome overduplication with polyploidy and multilobular nuclei in response to nocodazole37. This implicates p21 as one of the main p53 target genes in this context. Importantly, the overexpression of BCL2 allows polyploid p53-deficient cells that are generated by nocodazole treatment to survive more efficiently40. So, p53 abrogation and apoptosis inhibition can cooperate to induce rapid and progressive polyploidization following mitotic spindle damage. It has also been reported that p53-dependent apoptosis occurs in cells that undergo mitotic slippage and in aneuploid cells that are the result of aberrant multipolar mitoses, and that this process involves the increased expression of the death receptor CD95 and the cell-cycle regulators p21 and B-cell translocation gene-2 (BTG2)41. The absence of functional p53 is permissive for the survival of polyploid cells that can be formed in several ways. These include the downregulation of the PASSENGER 42 PROTEIN survivin by siRNA , treatment with spindle poisons such as nocodazole43, cell fusion with polyethylene glycol44, or endopolyploidization due to the expression

of the mitotic kinases Aurora-A, Aurora-B and pololike kinases45,46 or K-cyclin (the Kaposi’s sarcoma-associated herpesvirus cyclin-D homologue)47. Similarly, p53-knockout mice manifest the accumulation of multinuclear cells in various organs48. The inactivation of p53 (and Rb) by the SV40 large T-cell antigen can lead to tetraploidy in the mouse pancreas49 as well as to cytological abnormalities (polyploidy, ANISOCYTOSIS and ANISOKARYIOSIS) in the mouse liver, which precede hepatic carcinogenesis50. This latter effect is prevented by liverspecific overexpression of a p53 transgene50. The link between p53 abrogation and aneuploidy has also been established in a series of human cancers. For example, allelic losses of 17p (which contains the p53 gene) occur in diploid cells of patients with Barrett’s oesophagus before aneuploidy becomes manifest51. So, p53-dependent apoptosis could be an important mechanism that leads to the removal of polyploid cells, which constitute bona fide precursors of aneuploid cells. Mitotic cell death and genomic instability

Mitotic catastrophe. Cell death that occurs during metaphase (a phenomenon that is known as mitotic catastrophe) can be induced by DNA damage or by the fusion of asynchronous cells, provided that the DNA-STRUCTURE CHECKPOINT is inactivated (for example, by the inhibition of CHK2; REF. 52; FIG. 4). This type of cell death is independent of p53. Although morphologically distinct from apoptosis53, mitotic catastrophe does involve the activation of the apoptotic machinery. This includes signs of MMP, such as the loss of the mitochondrial transmembrane potential (∆Ψm), the mitochondrial release of CYTOCHROME C and APOPTOSIS-INDUCING FACTOR (AIF), caspase activation and DNA fragmentation54. In one model of mitotic catastrophe — the metaphaseassociated death of CHK2-inhibited HeLa SYNCYTIA — the order of lethal events has been established. Here, caspase-2, which is activated by DNA damage (see above), functions as an initiator caspase upstream of MMP, which, in turn, is required for the activation of the effector caspase, caspase-3 (REF. 55).

Box 4 | Cell fusion as a source of tumour-cell diversity Syncytium formation is a physiological process that is coupled to irreversible end-stage differentiation in the generation of the syncytiotrophoblast, muscle fibres, osteoclasts or foreign body giant cells (cells that are formed during local inflammation, in response to alien particles). Such syncytia are characterized by cytogamy (cytoplasmic fusion) without karyogamy (nuclear fusion) and therefore remain polynuclear (with all nuclei in the G0/G1 phase). Heterotropic, presumably non-physiological, cell fusion has been extensively documented in vivo, by experiments in which stem cells that bear certain genetic markers (such as Y chromosomes, major histocompatability complex (MHC) alleles or antibiotic-resistance genes) are transfused into recipient mice93–95. As the fusion products are mononuclear, they must result from a combination of cytogamy and karyogamy. Initially, this phenomenon has been misinterpreted as ‘transdifferentiation’ (the conversion of one cell type to another one). Similarly, tumour cells of human origin can fuse with host cells when inoculated into hamsters or mice96. And, in chimeric mice, carcinogens can result in the formation of tumours that bear markers from both parental strains97. It has been suggested that tumour cells are intrinsically more fusogenic than their normal counterparts98. Cell fusion might lead to the acquisition of chemotherapy resistance from both parental cells99, as well as to a temporary resistance to apoptosis98,100. Surviving hybrid cells carry chromosomal aberrations including chromosome dysfunctions, mitotic recombinations, deletions, insertions and inversions98. Moreover, cell fusion (for example, between melanoma cells and macrophages) can result in tumour cells with an enhanced metastatic potential101.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 5 | SEPTEMBER 2004 | 7 5 7 ©2004 Nature Publishing Group

REVIEWS

2N 2N

Cell fusion Endoreplication 2N

p53 Apoptosis

or

Defective cytoor karyokinesis

BCL2

Diploidy

4N

Polyploidy

2N (+/– nC)

Aneuploidy

Figure 3 | Relationship between polyploidy and apoptosis. Polyploidy can result from a variety of alterations; for example, as a result of cell fusion, endomitosis or failed cell division (BOX 1). Polyploid cells normally undergo apoptosis following the activation of a p53-dependent checkpoint. Inhibition of apoptosis, for example by overexpressed BCL2, favours multipolar and asymmetric divisions and therefore results in aneuploidy. nC, number of chromosomes.

CYTOCHROME C

A haem protein that is normally confined to the mitochondrial intermembrane space. Following induction of apoptosis, cytochrome c is released from mitochondria and triggers the formation of the apoptosome, a caspase-activation complex. APOPTOSIS-INDUCING FACTOR

(AIF). A flavoprotein that is normally present in the mitochondrial intermembrane space. Following the induction of apoptosis, AIF translocates to the nucleus where it activates a molecular complex that causes large-scale DNA fragmentation, presumably in a caspaseindependent fashion. SYNCYTIUM

A cytoplasm that contains several nuclei within the same plasma membrane and without internal cell boundaries. Syncytia are mostly generated by cell fusion. KARYOKINESIS

The physical separation of the daughter nuclei at the end of mitosis. SPINDLE-ASSEMBLY CHECKPOINT

A checkpoint that monitors the correct attachment of chromosomes to spindles in the metaphase–anaphase transition. Activation of this checkpoint causes cell-cycle arrest as a result of the inhibition of anaphasepromoting complex/cyclosome (APC/C) and is mediated by a cytoplasmic activity known as cytostatic factor (CSF).

758

Importantly, the inhibition of apoptosis by the suppression of caspases or MMP prevents mitotic catastrophe55. In conditions of disabled apoptosis, cells can continue mitosis beyond the metaphase and complete both KARYOKINESIS and cytokinesis. Fluorescent videomicroscopy of syncytia — in which chromosomes have been labelled with a histone H2A–GFP (green fluorescent protein) fusion protein — shows that, in conditions of apoptosis suppression, cells can simultaneously divide into three or more heterogeneous daughter cells. This cellular heterogeneity is manifest in terms of both cell size (anisocytosis) and DNA content (anisokaryosis). Most of the cells that are generated by asymmetric division are aneuploid55. Similar data have been reported for a model of DNA damage in which the simultaneous abolition of the DNA-structure checkpoint and inhibition of apoptosis leads to rapid aneuploidization55. So, mitotic catastrophe emerges as a special case of metaphase-associated apoptosis, the suppression of which leads to asymmetric division and aneuploidy. The spindle-assembly checkpoint. The SPINDLE-ASSEMBLY CHECKPOINT delays mitotic progression until all chromosomes have achieved a proper bipolar orientation on the mitotic spindle. The checkpoint is triggered by KINETOCHORES that lack attached microtubules and/or by a lack of tension in the spindle46. A defective spindle checkpoint or imprecise sister-chromatid separation are associated with numeric CIN. So, mutations in BUB1, MAD1, MAD2, BUBR1 (a homologue of yeast Mad3) or securin, which are all involved in these processes, can participate in the induction of cancerassociated CIN56. The expression of BUBR1, a mitotic checkpoint kinase, is significantly reduced in polyploid cells that were created by sustained damage to the mitotic spindle (for example, by culturing cells in the presence of nocodazole), as well as in a significant fraction of colon carcinomas. The reduction of BUBR1 expression might be explained by increased ubiquitindependent proteolysis in polyploid cells that arises during prolonged mitotic arrest57. Re-introduction of BUBR1 triggered the apoptosis of polyploid cells that

| SEPTEMBER 2004 | VOLUME 5

are formed by aberrant exit from mitosis, and inhibited the growth of human adenocarcinomas that were transplanted into nude mice57. The mitotic kinase Aurora-B and its associated passenger proteins INCENP and survivin participate in the tension-activated arm of the spindle-assembly checkpoint. Knockdown or pharmacological inhibition of either of these proteins can cause defects in chromosome segregation and cytokinesis46,58. Moreover, knockdown of survivin expression can cause several centrosomal defects, aberrant multipolar spindle formation and chromatin missegregation, and these phenotypes are exacerbated in p53 –/– and p21 –/– cells42. The knockdown of survivin expression also causes p53-independent MMP and apoptosis42, a finding that has been confirmed by the overexpression of a dominant-negative survivin mutant, which triggers MMP and the rapid release of AIF and cytochrome c 59. This points to an important degree of networking between the spindle checkpoint and apoptosis regulation. It remains to be determined how survivin suppresses apoptosis — whether it is by a direct inhibitory effect on the mitochondrial caspase activator SMAC/DIABLO60 or via a ternary interaction with two other proteins, HBXPIP (hepatitis-B X-interacting protein) and pro-caspase-9, which would also result in caspase inhibition61. Apoptosis versus cell-cycle regulation by p53

As mentioned above, p53 is frequently required for the suppression of DSBs, ‘naked’ telomeres, polyploid genomes or defective mitoses. Accordingly, p53-null cell lines or tumours frequently become aneuploid. It has been a matter of intense debate as to whether the absence of p53 induces aneuploidy through a failure of the cells to arrest the cell cycle or through deficient apoptosis. Myc-induced lymphomas that arise in EµMyc-transgenic mice develop in an accelerated fashion, both in the context of p53 deficiency or in the presence of an additional BCL2 transgene. However, p53-null lymphomas, which showed checkpoint defects and were highly aneuploid, differed from BCL2-expressing lymphomas (that have intact p53), which retained intact checkpoints and were largely diploid. This has been used as an argument to say that the loss of p53 in murine lymphomas causes genomic instability through mechanisms that are not related to apoptosis62,63. In accordance with this particular interpretation, a knockin mutation of p53, which maintained the p53-mediated G1-phase arrest (through its capacity to induce p21) but abolished the pro-apoptotic function of p53, reduced the frequency of thymic lymphomas as well as the degree of aneuploidy found in such tumours, compared with p53-knockout mice64. On the basis of these latter results, it has been postulated that cell-cycle arrest, rather than apoptosis induction, might mediate the genome-stabilizing action of p53 (REF. 64). However, this debate is ongoing. First, p21 is now thought to have a second, apoptosis-inhibitory function65, thereby raising doubts about the clear distinction between cell-cycle regulation and apoptosis induction by p53. Second, some data indicate that, in

www.nature.com/reviews/molcellbio ©2004 Nature Publishing Group

REVIEWS

p53 Apoptosis CHK2

Cell-cycle arrest at G2/M

CDK1

Metaphase

CDK1

Asynchronous cell fusion

Mitotic prophase

Apoptosis (mitotic catastrophe)

Asymmetric cell division

Figure 4 | Mitotic catastrophe as a default pathway to avoid asymmetric division. Fusion of asynchronous cells results in a CHK2-dependent prophase arrest, which eventually activates p53 and apoptosis, as a default pathway. The inactivation of CHK2 allows entry into metaphase, at which point cells undergo mitotic catastrophe unless the apoptotic machinery is disabled, for example, due to the inhibition of caspases or expression of BCL2.

KINETOCHORE

A specialized condensed chromosomal region in which the chromatids are held together to form an X shape. SMAC/DIABLO

A mitochondrial intermembrane protein that, on apoptotic release, can interact with inhibitor of apoptosis (IAP) proteins such as XIAP, thereby inhibiting their function and facilitating caspase activation. BASE-EXCISION REPAIR

(BER). The main pathway that is responsible for the repair of apurinic and apyrimidinic (AP) sites in DNA. BER is catalyzed in four consecutive steps by a DNA glycosylase, which removes the damaged base; an AP endonuclease, which processes the abasic site; a DNA polymerase, which inserts the new nucleotide(s); and DNA ligase, which rejoins the DNA strand. NUCLEOTIDE-EXCISION REPAIR

(NER). A process in which a small region of the DNA strand that surrounds DNA damage is removed from the DNA helix as an oligonucleotide.

model organisms such as Drosophila melanogaster, the pro-death role of p53 prevails over its anti-proliferative action in assuring genomic instability 66. The elimination of the D. melanogaster p53 homologue (Dmp53) causes genomic instability, radiation hypersensitivity, apoptosis defects (and the reduced expression of the apoptosis effectors Reaper and Sickle in response to DNA damage), yet has no effects on DNA-damage-induced cellcycle arrest66. Finally, the genome-stabilizing and tumour-suppressive effects of p53 might be regulated differentially. In MYC-induced lymphomas, which develop under the influence of a BCL2 transgene, there is no selective pressure for losing p53 expression. This indicates that the principal anti-tumour role of p53 is in inducing apoptosis rather than preventing genomic instability 62. This is at odds with the knock-in study 64, according to which the cell-cycle-inhibitory function of p53 (which also represses genomic instability) would suffice for tumour suppression. Future studies will have to address the relationship between p53, genomic instability, tumour development and cancer treatment in defined (rather than several distinct) experimental systems, so that the contribution of p53-mediated cellcycle control versus apoptosis induction can be weighed. Crosstalk between apoptosis and DNA repair

There are three scenarios for how apoptosis might relate to DNA repair. First, proteins that detect DNA damage can directly relay to the apoptotic machinery (see above). Second, apoptosis regulators such as BCL2 can participate in the regulation of DNA repair. Third, oncogenic kinases can simultaneously inhibit apoptosis and DNA repair. There are numerous examples of proteins that participate in the detection of DNA damage and in DNA repair and that also can stimulate apoptosis, as a default pathway. The poly(ADP-ribose) polymerase-1 and -2 (PARP-1, PARP-2) can participate in BASE-EXCISION REPAIR (BER), as well as in signalling pathways that lead to

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

apoptosis. When highly activated, PARP causes the depletion of nicotine adenine dinucleotide (NAD) and/or the accumulation of poly(ADP-ribose), which in turn stimulates MMP and the release of AIF, thereby triggering caspase-independent apoptosis67,68. Many proteins in the BRCA1-associated genome surveillance complex (BASC) — which contains the breast-cancerassociated protein BRCA1, the Bloom syndrome DNA helicase BLM, mismatch-repair proteins (MSH2, MSH6, MLH1), replication factor C (RFC) and ATM — also participate both in DNA repair and in signalling DNA damage to p53. Consistent with this idea, the stimulation of p53 phosphorylation, which is induced by DNA-methylation damage, is dependent on functional MSH and MLH mismatch-repair proteins that are capable of recognizing O6-methylguanine69. The apoptotic response that is elicited by O6-methylguanine is mainly mediated by the intrinsic, mitochondrial pathway of apoptosis70 (BOX 2). The BLM and the Werner syndrome (WRN) DNA helicases are involved in the unwinding of intermediates of recombination, thereby preventing uncontrolled recombination events. Their associated deficiencies give rise to elevated levels of recombination (the hyperrecombination phenotype), which result in chromosomal aberrations, including the loss of heterozygosity. Importantly, BLM and WRN can bind to p53, and BLM and WRN deficiencies attenuate p53-mediated apoptosis following DNA damage71,72. High levels of chromosome aberrations after treatment with ionizing radiation have been reported in human lymphoblast cell lines that express the antiapoptotic regulator BCL2 (REF. 73). Similarly, BCL2 overexpression can increase the rate of mutagenesis that is induced by genotoxic stress, such as the oxidative stress that is caused by benzene metabolites. This correlates with a reduced removal of damaged bases (8-hydroxydeoxyguanosine and thymol glycol) from the DNA of the cell74. BCL2 overexpression (and that of the related BCL2-family member BCL-XL) can reduce the expression of functional RAD51 protein, thereby reducing RAD51-dependent DNA repair75. As this effect has also been found for the BCL2 mutant in which Gly145 is mutated to Ala, and which has lost its anti-apoptotic potential, it has been argued that anti-apoptotic proteins of the BCL2 family favour a mutator phenotype that is due to the reduction of error-free DNA repair, and that this effect is independent of its anti-apoptotic activity75. Oncogenic tyrosine kinases can combine anti-apoptotic and anti-repair effects. The constitutively active BCR–ABL kinase (which arises from the fusion of the kinase domain of ABL with the BCR protein) inhibits apoptosis through the transcriptional activation of BCL-XL and inactivates the pro-apoptotic regulator BAD, in part by its direct phosphorylation. Overexpression of the p210 BCR–ABL isoform modulates NUCLEOTIDE-EXCISION REPAIR (NER) in a lineage-specific fashion; it enhances NER in myeloid cells and reduces NER in lymphoid cells76. Moreover, BCR–ABL might stimulate DSB repair by homologous recombination, presumably due to the signal transducer and

VOLUME 5 | SEPTEMBER 2004 | 7 5 9 ©2004 Nature Publishing Group

REVIEWS activator of transcription (STAT)5-mediated induction of RAD51 (REF. 77). By contrast, BCR–ABL reportedly downregulates DNA-PKcs, which is involved in NHEJ 78. These phenomena might participate in the resistance of chronic myeloid leukaemia cells to genotoxic therapies. It has been speculated that the reduced NHEJ and the enhanced homologous-recombination-repair activities might be responsible for intra-chromosomal or interchromosomal deletions and chromosomal translocations that are observed in BCR–ABL-positive leukaemias79. Another example is provided by the oncogenic tyrosine kinase LCK, which inhibits DSB repair as well as the DNA-damage-induced BCL-XL deamidation, a post-transcriptional modification that inactivates BCL-XL (REF. 80). This process might well be involved in T-cell transformation (in the case of LCK), as well as in transformation that is mediated by other oncogenic tyrosine kinases. Altogether, these findings point to an intricate interplay between regulators and effectors of the DNA-repair and apoptosis-execution machineries. It seems that deficient DNA repair can be coupled to a failure to elicit an apoptotic response, and this particular association could participate in oncogenesis. Conclusions and perspectives

As discussed here, the machinery that controls and executes cell death and the mechanisms that stabilize or endanger the accurate replication of the genome engage in an intricate interplay, the detailed comprehension of which is still in its infancy. However, a few rules have emerged that control the relationship between apoptosis and genome maintenance. Inhibition of apoptosis can favour CIN at several levels. So, DSBs cause structural CIN when the default pathway that leads to senescence and apoptosis is blocked. Similarly, telomere dysfunction entails rampant structural CIN only when the p53dependent senescence or apoptosis pathway is disabled. Inactivation of p53 is also permissive for the survival of polyploid cells. Disabling the DNA-structure checkpoint can favour metaphase-associated death. But suppression of this mitotic catastrophe by caspase inhibitors or BCL2 overexpression results in asymmetric division and aneuploidy. Furthermore, the spindle-assembly checkpoint is functionally linked to apoptosis regulation by the proapoptotic BUBR1 and anti-apoptotic survivin proteins. Oncogenic kinases can simultaneously inhibit DNA

1.

2. 3.

760

Hanahan, D. & Weinberg, R. A. The hallmarks of cancer. Cell 100, 57–70 (2000). Summarizes essential alterations in cell physiology, which dictate malignant transformation. Pihan, G. & Doxsey, S. J. Mutations and aneuploidy: co-conspirators in cancer? Cancer Cell 4, 89–94 (2003). Ionov, Y., Yamamoto, H., Krajewski, S., Reed, J. C. & Perucho, M. Mutational inactivation of the proapoptotic gene BAX confers selective advantage during tumor clonal evolution. Proc. Natl Acad. Sci. USA 97, 10872–10877 (2000). First evidence that genomic instability of the microsatellite type can induce mutations in apoptosis-regulatory genes, thereby providing a growth advantage to tumours in vivo. This study is backed up by clinical data that validates the concept that BAX is frequently mutated in colon carcinomas.

4.

5.

6.

7. 8.

repair and apoptosis. Moreover, proteins that are involved in DNA repair can participate in the activation of the DNA-damage-induced apoptotic default pathway, or, vice versa, apoptosis-regulatory proteins can affect DNA repair. Future research must identify the detailed mechanistic links between these phenomena, at the biochemical and genetic levels. In the context of oncogenesis, it is an unresolved ‘chicken and egg’ question how apoptosis resistance and genomic instability are related in chronological and mechanistic terms, especially in the early phases of cancer development. It can be expected (but still has to be proven) that a significant fraction of cells with scrambled genomes, once freshly generated through BFB cycles or asymmetric cell divisions, will manifest a lethal metabolic insufficiency or decompensating signalling network that might favour cellular demise. Although most of the potentially malignant cells would eventually succumb, the apoptotic threshold (as set, for example, by the expression level of pro- and anti-apoptotic BCL2-family proteins or that of caspases and their endogenous inhibitors) would determine the probability that such cells can survive and eventually adapt to new niches of the host organism. In this theoretical scenario, the development of malignancy would result from the survival of selected clones that carry complex, cancer-specific alterations in the genome as well as in cellular functions, including (relative) apoptosis resistance. So, genomic instability might relate to disabled apoptosis following a series of distinct patterns, and both phenomena — instability and suppressed death — could engage in a positive amplification loop. Far from being a merely academic playground, the intricate interplay between apoptosis control and genomic instability is likely to have an important role in oncogenesis and therefore constitutes a challenge for applied biomedical research. As discussed in this review, proteins that determine this interplay include p53, cell-cycle regulators such as p21 and checkpoint kinases such as CHK2, as well as numerous additional effectors that are involved in the DNA-damage and spindleassembly checkpoints. Although it is almost certain that this theatre will be joined by additional actors, it might constitute an initial panel of targets for the prevention of cancer development, as well as for stopping deleterious genetic drifts in established malignancies before they become incurable.

Gourdier, I. et al. Drug specific resistance to oxaliplatin is associated with apoptosis defect in a cellular model of colon carcinoma. FEBS Lett. 529, 232–236 (2002). Reid, B. J. et al. Predictors of progression in Barrett’s esophagus II: baseline 17p (p53) loss of heterozygosity identifies a patient subset at increased risk for neoplastic progression. Am. J. Gastroenterol. 96, 2839–2848 (2001). Sugai, T. et al. Analysis of genetic alterations, classified according to their DNA ploidy pattern, in the progression of colorectal adenomas and early colorectal carcinomas. J. Pathol. 200, 168–176 (2003). Norbury, C. J. & Zhitovovsky, B. DNA damage-induced apoptosis. Oncogene 23, 2797–2808 (2004). Shiloh, Y. ATM and related protein kinases: safeguarding genome integrity. Nature Rev. Cancer 3, 155–168 (2003).

| SEPTEMBER 2004 | VOLUME 5

9.

10.

11.

12.

13.

Bermudez, V. P. et al. Loading of the human 9-1-1 checkpoint complex onto DNA by the checkpoint clamp loader hRad17-replication factor C complex in vitro. Proc. Natl Acad. Sci. USA 100, 1633–1638 (2003). Motoyama, N. & Naka, K. DNA damage tumor suppressor genes and genomic instability. Curr. Opin. Genet. Dev. 14, 11–16 (2004). Peters, M. et al. Chk2 regulates irradiation-induced, p53-mediated apoptosis in Drosophila. Proc. Natl Acad. Sci. USA 99, 11305–11310 (2002). Takai, H. et al. Chk2-deficient mice exhibit radioresistance and defective p53-mediated transcription. EMBO J. 21, 5195–5205 (2002). Celeste, A. et al. Genomic instability in mice lacking histone H2AX. Science 296, 922–927 (2002). Pioneering work in which the role for histone H2AX in genomic instability was first proposed.

www.nature.com/reviews/molcellbio ©2004 Nature Publishing Group

REVIEWS 14. Taneja, N. et al. Histone H2AX phosphorylation as a predictor of radiosensitivity and target for radiotherapy. J. Biol. Chem. 279, 2273–2280 (2004). 15. Suzuki, K., Yokoyama, S., Waseda, S., Kodama, S. & Watanabe, M. Delayed reactivation of p53 in the progeny of cells surviving ionizing radiation. Cancer Res. 63, 936–941 (2003). 16. Vousden, K. H. & Lu, X. Live or let die: the cell’s response to p53. Nature Rev. Cancer 2, 594–604 (2002). 17. Konishi, A. et al. Involvement of histone H1.2 in apoptosis induced by DNA double-strand breaks. Cell 114, 673–688 (2003). 18. Chipuk, J. E., Maurer, U., Green, D. R. & Schuler, M. Pharmacologic activation of p53 elicits Bax-dependent apoptosis in the absence of transcription. Cancer Cell 4, 371–381 (2003). Shows convincing evidence that p53-dependent apoptosis might proceed in a transcriptionally independent manner. 19. Marchenko, N. D., Zaika, A. & Moll, U. M. Death signalinduced localization of p53 protein to mitochondria. A potential role in apoptotic signaling. J. Biol. Chem. 275, 16202–16212 (2000). 20. Mihara, M. et al. p53 has a direct apoptogenic role at the mitochondria. Mol. Cell 11, 577–590 (2003). References 19 and 20 show the possible transcriptionindependent effect of p53 on mitochondria, whereby p53 would interact with BCL-XL. 21. Chipuk, J. E. et al. Direct activation of Bax by p53 mediates mitochondrial membrane permeabilization and apoptosis. Science 303, 1010–1014 (2004). 22. Leu, J. I.-J., Dumont, P., Hafey, M., Murphy, M. E. & George, D. L. Mitochondrial p53 activates BAK and leads to disruption of a BAK–MCL1 complex. Nature Cell Biol. 6, 443–450 (2004). References 21 and 22 indicate that the true mitochondrial target of p53 would be BAX (which would be activated in a hit-and-run fashion) or MCL1, respectively. 23. Melino, G. et al. p73 induces apoptosis via PUMA transactivation and Bax mitochondrial translocation. J. Biol. Chem. 279, 8076–8083 (2004). 24. Lin, B. et al. Conversion of Bcl-2 from protector to killer by interaction with nuclear orphan receptor Nur77/TR3. Cell 116, 527–540 (2004). 25. Tinel, A. & Tschopp, J. The PIDDosome, a protein complex implicated in activation of caspase-2 in response to genotoxic stress. Science 304, 843–846 (2004). The description of a nuclear caspase-activation complex — the PIDDosome — that contains PIDD (which is also a p53 target gene), the adaptor protein RAIDD/CRAIDD and pro-caspase-2. 26. Ben-Porath, I. & Weinberg, R. A. When cells get stressed: an integrative view of cellular senescence. J. Clin. Invest. 113, 8–13 (2004). 27. Artandi, S. E. et al. Constitutive telomerase expression promotes mammary carcinomas in aging mice. Proc. Natl Acad. Sci. USA 99, 8191–8196 (2002). 28. Gisselsson, D. Chromosome instability in cancer: how, when, and why? Adv. Cancer Res. 87, 1–29 (2003). 29. DePinho, R. A. The age of cancer. Nature 408, 248–254 (2000). Proposes the theory that carcinogenesis is first triggered by a crisis with telomere dysfuction and genomic instability, followed by re-expression of telomerase. 30. Wong, K. K. et al. Telomere dysfunction and Atm deficiency compromises organ homeostasis and accelerates ageing. Nature 421, 643–648 (2003). Reports in vivo evidence that telomere dysfunction, AMT and p53 participate in the regulation of CIN, for example, in the crypts of colon epithelia where cancer stem cells are expected to form. 31. Chin, L. et al. p53 deficiency rescues the adverse effects of telomere loss and cooperates with telomere dysfunction to accelerate carcinogenesis. Cell 97, 527–538 (1999). 32. Zhang, P., Chan, S. L., Fu, W., Mendoza, M. & Mattson, M. P. TERT suppresses apoptotis at a premitochondrial step by a mechanism requiring reverse transcriptase activity and 14-3-3 protein-binding ability. FASEB J. 17, 767–769 (2003). 33. Kang, H. J. et al. Ectopic expression of the catalytic subunit of telomerase protects against brain injury resulting from ischemia and NMDA-induced neurotoxicity. J. Neurosci. 24, 1280–1287 (2004). References 32 and 33 suggest an apoptosis-inhibitory function of the protein subunit of telomerase that might be dissociated from its effect on telomeres. 34. Storchova, Z. & Pellman, D. From polyploidy to aneuploidy, genome instability and cancer. Nature Rev. Mol. Cell Biol. 5, 45–54 (2004).

35. Raslova, H. et al. Megakaryocyte polyploidization is associated with functional gene amplification. Blood 101, 541–544 (2003). 36. Guidotti, J. E. et al. Liver cell polyploidization: a pivotal role for binuclear hepatocytes. J. Biol. Chem. 278, 19095–19101 (2003). 37. Mantel, C. et al. p21cip-1/waf-1 deficiency causes deformed nuclear architecture, centriole overduplication, polyploidy, and relaxed microtubule damage checkpoints in human hematopoietic cells. Blood 93, 1390–1398 (1999). 38. Lanni, J. S. & Jacks, T. Characterization of the p53-dependent postmitotic checkpoint following spindle disruption. Mol. Cell Biol. 18, 1055–1064 (1998). 39. Andreassen, P. R., Lacroix, F. B., Lohez, O. D. & Margolis, R. L. Neither p21WAF1 nor 14-3-3σ prevents G2 progression to mitotic catastrophe in human colon carcinoma cells after DNA damage, but p21WAF1 induces stable G1 arrest in resulting tetraploid cells. Cancer Res. 61, 7660–7668 (2001). 40. Minn, A. J., Boise, L. H. & Thompson, C. B. Expression of Bcl-xL and loss of p53 can cooperate to overcome a cell cycle checkpoint induced by mitotic spindle damage. Genes Dev. 10, 2621–2631 (1996). Established the role of p53 and BCL2-family proteins in the regulation of the cell-cycle checkpoint following the induction of damage to the mitotic spindle. 41. Chen, J. G., Yang, C. P., Cammer, M. & Horwitz, S. B. Gene expression and mitotic exit induced by microtubulestabilizing drugs. Cancer Res. 63, 7891–7899 (2003). 42. Beltrami, E., Plescia, J., Wilkinson, J. C., Duckett, C. S. & Altieri, D. C. Acute ablation of survivin uncovers p53-dependent mitotic checkpoint functions and control of mitochondrial apoptosis. J. Biol. Chem. 279, 2077–2084 (2004). 43. Verdoodt, B. et al. Induction of polyploidy and apoptosis after exposure to high concentrations of the spindle poison nocodazole. Mutagenesis 14, 513–520 (1999). 44. Castedo, M. et al. Human immunodeficiency virus 1 envelope glycoprotein complex-induced apoptosis involves mammalian target of rapamycin/FKBP12-rapamycinassociated protein-mediated p53 phosphorylation. J. Exp. Med. 194, 1097–1110 (2001). 45. Meraldi, P., Honda, R. & Nigg, E. A. Aurora-A overexpression reveals tetraploidization as a major route to centrosome amplification in p53 –/– cells. EMBO J. 21, 483–492 (2002). A seminal report that outlines how centrosome amplification could lead to aneuploidy in a p53-negative context. 46. Carmena, M. & Earnshaw, W. C. The cellular geography of aurora kinases. Nature Rev. Mol. Cell Biol. 4, 842–854 (2003). 47. Verschuren, E. W., Klefstrom, J., Evan, G. I. & Jones, N. The oncogenic potential of Kaposi’s sarcoma-associated herpesvirus cyclin is exposed by p53 loss in vitro and in vivo. Cancer Cell 2, 229–241 (2002). 48. Colombel, M. et al. Androgen suppressed apoptosis is modified in p53 deficient mice. Oncogene 10, 1269–1274 (1995). 49. Ramel, S. et al. Inactivation of p53 and the development of tetraploidy in the elastase-SV40 T antigen transgenic mouse pancreas. Pancreas 11, 213–222 (1995). 50. Gillet, R. et al. The consequence of p53 overexpression for liver tumor development and the response of transformed murine hepatocytes to genotoxic agents. Oncogene 19, 3498–3507 (2000). 51. Blount, P. L. et al. 17p allelic losses in diploid cells of patients with Barrett’s esophagus who develop aneuploidy. Cancer Res. 54, 2292–2295 (1994). 52. Castedo, M. et al. Chk2 is a negative regulator of mitotic catastrophe. Oncogene 23, 4353–4361 (2004). 53. Roninson, I. B., Broude, E. V. & Chang, B.-D. If not apoptosis, then what? Treatment-induced senescence and mitotic catastrophe in tumor cells. Drug Resist. Updat. 4, 303–313 (2001). 54. Castedo, M., Perfettini, J.-L., Medema, J. P. & Kroemer, G. Cell death by mitotic catastrophe: a molecular definition. Oncogene 23, 2825–2837 (2004). 55. Castedo, M. et al. Mitotic catastrophe. A special case of apoptosis preventing aneuploidy. Oncogene 23, 4362–4370 (2004). The first systematic investigation of biochemical and morphological events that are related to mitotic catastrophe. 56. Jallepalli, P. V. & Lengaue, C. Chromosome segregation and cancer: cutting through the mystery. Nature Rev. Cancer 1, 109–117 (2001). 57. Shin, H. J. et al. Dual roles of human BubR1, a mitotic checkpoint kinase, in the monitoring of chromosomal instability. Cancer Cell 4, 483–497 (2003).

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

58. Lens, S. M. et al. Survivin is required for a sustained spindle checkpoint arrest in response to lack of tension. EMBO J. 22, 2934–2947 (2003). Outlines the role of survivin in the tension arm of the spindle checkpoint. 59. Liu, T., Brouha, B. & Grossman, D. Rapid induction of mitochondrial events and caspase-independent apoptosis in Survivin-targeted melanoma cells. Oncogene 23, 39–48 (2004). 60. Song, Z., Yao, X. & Wu, M. Direct interaction between survivin and Smac/DIABLO is essential for the anti-apoptotic activity of survivin during taxol-induced apoptosis. J. Biol. Chem. 278, 23130–23140 (2003). 61. Marusawa, H. et al. HBXIP functions as a cofactor of survivin in apoptosis suppression. EMBO J. 22, 2729–2740 (2003). 62. Schmitt, C. A. et al. Dissecting p53 tumor suppressor functions in vivo. Cancer Cell. 1, 289–298 (2002). An important study in which the contribution of p53 inactivation to lymphomagenesis is assessed in vivo, either in the absence or in the presence of a BCL2 transgene, which actually alleviates the pressure to inactivate p53. 63. Fridman, J. S. & Lowe, S. W. Control of apoptosis by p53. Oncogene 22, 9030–9040 (2003). 64. Liu, G. et al. Chromosome stability, in the absence of apoptosis, is critical for suppression of tumorigenesis in Trp53 mutant mice. Nature Genet. 36, 63–68 (2004). A study in which a knock-in mutation of p53, which is supposed to have little or no apoptosis-inducing potential, is examined for its tumour-suppressive effect. 65. Lawson, B. R. et al. Deficiency of the cyclin kinase inhibitor p21WAF-1/CIP-1 promotes apoptosis of activated/memory T cells and inhibits spontaneous systemic autoimmunity. J. Exp. Med. 199, 547–557 (2004). 66. Sogame, N., Kim, M. & Abrams, J. M. Drosophila p53 preserves genomic stability by regulating cell death. Proc. Natl Acad. Sci. USA 100, 4696–4701 (2003). 67. Cregan, S. P. et al. Apoptosis-inducing factor is involved in the regulation of caspase-independent neuronal cell death. J. Cell Biol. 158, 507–517 (2002). 68. Alano, C. C., Ying, W. & Swanson, R. A. Poly(ADP-ribose) polymerase-1 mediated cell death in astrocytes requires NAD+ depletion and mitochondrial permeability transition. J. Biol. Chem. 279, 18895–18902 (2004). 69. Duckett, D. R., Bronstein, S. M., Taya, Y. & Modrich, P. hMutSα- and hMutLα-dependent phosphorylation of p53 in response to DNA methylator damage. Proc. Natl Acad. Sci. USA 96, 12384–12388 (1999). 70. Hickman, M. J. & Samson, L. D. Apoptotic signaling in response to a single type of DNA lesion, O6-methylguanine. Mol. Cell 14, 105–116 (2004). An exhaustive analysis of the molecular mechanisms that link oxidative DNA damage to activation of the apoptotic default pathway, thereby revealing crosstalk between detection of DNA damage by the DNArepair machinery and the activation of lethal signals. 71. Nakayama, H. RecQ family helicases: roles as tumor suppressor proteins. Oncogene 21, 9008–9021 (2002). 72. Davalos, A. R. & Campisi, J. Bloom syndrome cells undergo p53-dependent apoptosis and delayed assembly of BRCA1 and NBS1 repair complexes at stalled replication forks. J. Cell Biol. 162, 1197–1209 (2003). 73. Cherbonnel-Lasserre, C., Gauny, S. & Kronenberg, A. Suppression of apoptosis by Bcl-2 or Bcl-xL promotes susceptibility to mutagenesis. Oncogene 13, 1489–1497 (1996). 74. Kuo, M. L., Shiah, S. G., Wang, C. J. & Chuang, S. E. Suppression of apoptosis by Bcl-2 to enhance benzene metabolites-induced oxidative DNA damage and mutagenesis: a possible mechanism of carcinogenesis. Mol. Pharmacol. 55, 894–901 (1999). 75. Saintigny, Y. et al. Characterization of homologous recombination induced by replication inhibition in mammalian cells. EMBO J. 20, 3861–3870 (2001). 76. Canitrot, Y. et al. p210 BCR/ABL kinase regulates nucleotide excision repair (NER) and resistance to UV radiation. Blood 102, 2632–2637 (2003). 77. Slupianek, A. et al. BCR/ABL regulates mammalian RecA homologs, resulting in drug resistance. Mol. Cell 8, 795–806 (2001). 78. Deutsch, E. et al. BCR–ABL down-regulates the DNA repair protein DNA-PKcs. Blood 97, 2084–2090 (2001). 79. Skorski, T. BCR/ABL regulates response to DNA damage: the role in resistance to genotoxic treatment and in genomic instability. Oncogene 21, 8591–8604 (2002). 80. Zhao, R., Yang, F. T. & Alexander, D. R. An oncogenic tyrosine kinase inhibits DNA repair and DNA-damageinduced Bcl-xL deamidation in T cell transformation. Cancer Cell 5, 37–49 (2004).

VOLUME 5 | SEPTEMBER 2004 | 7 6 1 ©2004 Nature Publishing Group

REVIEWS

81.

82. 83. 84. 85.

86.

87. 88.

89.

90.

762

Description of LCK as a kinase that can simultaneously mediate the inhibition of DNA repair, DNA-damage-induced BCL-XL deamidation, BAX conformational change and mitochondrial translocation, cytochrome-c release and the apoptotic caspase execution cascade. Kerr, J. F. R., Wyllie, A. H. & Currie, A. R. Apoptosis: a basic biological phenomenon with wide-ranging implications in tissue kinetics. Br. J. Cancer 26, 239–257 (1972). Green, D. R. & Reed, J. C. Mitochondria and apoptosis. Science 281, 1309–1312 (1998). Wang, X. The expanding role of mitochondria in apoptosis. Genes Dev. 15, 2922–2933 (2002). Adams, J. M. Ways of dying: multiple pathways to apoptosis. Genes Dev. 17, 2481–2495 (2003). Zamzami, N. & Kroemer, G. Mitochondria in apoptosis. How Pandora’s box opens. Nature Rev. Mol. Cell. Biol. 2, 67–71 (2001). Igney, F. H. & Krammer, P. H. Death and anti-death: tumour resistance to apoptosis. Nature Rev. Cancer 2, 277–288 (2002). Hayflick, L. The limited in vitro lifetime of human diploid cell strains. Exp. Cell Res. 37, 614–636 (1965). Harley, C. B., Futcher, A. B. & Greider, C. W. Telomeres shorten during ageing of human fibroblasts. Nature 345, 458–460 (1990). Yeo, E. J., Hwang, Y. C., Kang, C. M., Choy, H. E. & Park, S. C. Reduction of UV-induced cell death in the human senescent fibroblasts. Mol. Cell 10, 415–422 (2000). Seluanov, A. et al. Change of the death pathway in senescent human fibroblasts in response to DNA damage is

91.

92. 93.

94.

95.

96.

97.

98.

caused by an inability to stabilize p53. Mol. Cell Biol. 21, 1552–1564 (2001). Schmitt, C. A. et al. A senescence program controlled by p53 and p16INK4a contributes to the outcome of cancer therapy. Cell 109, 335–346 (2002). Seminal paper in which premature or inducible senescence, in addition to apoptosis, were identified as a drug-responsive effector programme of cancer therapy. Tumours that have defects in senescencecontrolling genes, such as p53 or INK4A, display an impaired treatment response. Blagosklonny, M. V. Cell senescence and hypermitogenic arrest. EMBO Rep. 4, 358–362 (2003). Terada, N. et al. Bone marrow cells adopt the phenotype of other cells by spontaneous cell fusion. Nature 416, 542–545 (2002). Vassilopoulos, G., Wang, P. R. & Russell, D. W. Transplanted bone marrow regenerates liver by cell fusion. Nature 422, 901–904 (2003). Wang, X. et al. Cell fusion is the principal source of bonemarrow-derived hepatocytes. Nature 422, 897–901 (2003). References 94 and 95 show that cell fusion can occur in vivo among normal somatic cells, presumably in physiological conditions. Goldenberg, D. M., Pavia, R. A. & Tsao, M. C. In vivo hybridisation of human tumour and normal hamster cells. Nature 250, 649–651 (1974). Fortuna, M. B., Dewey, M. J. & Furmanski, P. Enhanced lung colonization and tumorigenicity of fused cells isolated from primary MCA tumors. Cancer Lett. 55, 109–114 (1990). Duelli, D. & Lazebnik, Y. Cell fusion: a hidden enemy? Cancer Cell 3, 445–448 (2003).

| SEPTEMBER 2004 | VOLUME 5

An excellent review that summarizes the evidence on cell fusion as a putative oncogenic event. 99. Miller, F. R., Mohamed, A. N. & McEachern, D. Production of a more aggressive tumor cell variant by spontaneous fusion of two mouse tumor subpopulations. Cancer Res. 49, 4316–4321 (1989). 100. Duelli, D. M. & Lazebnik, Y. A. Primary cells suppress oncogene-dependent apoptosis. Nature Cell Biol. 2, 859–862 (2000). 101. Pawelek, J. M. Tumour cell hybridization and metastasis revisited. Melanoma Res. 10, 507–514 (2000).

Acknowledgements The authors’ own work is supported by the European Commission.

Competing interests statement The authors declare no competing financial interests.

Online links DATABASES The following terms in this article are linked online to: Entrez: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi MSH2 | p53 | TERC Swiss-Prot: http://us.expasy.org/sprot/ AIF | ATM | Aurora-A | Aurora-B | BAK | BCL2 | BCL-XL | BLM | BRCA1 | caspase-2 | CD95 | CHK1 | CHK2 | cytochrome c | H2AX | histone H1.2 | NUR77 | p21 | p73 | RAD1 | survivin | TERT | WRN Access to this links box is available online.

www.nature.com/reviews/molcellbio ©2004 Nature Publishing Group