Senescence from G2 arrest, revisited

Abstract

Senescence was classically defined as an irreversible cell cycle arrest in G1 phase (G1 exit) triggered by eroded telomeres in aged primary cells. The molecular basis of this G1 arrest is thought to be due to a DNA damage response, resulting in accumulation of the cyclin dependent kinase (Cdk) inhibitors p21 and p16 that block the inactivating phosphorylation of the retinoblastoma tumor suppressor pRb, thereby preventing DNA replication. More than a decade ago, several studies showed that p21 also mediates permanent DNA damage-induced cell cycle arrest in G2 (G2 exit) by inhibiting mitotic Cdk complexes and pRb phosphorylation. The idea that the senescence program can also be launched after G2 arrest has gained support from several recent publications, including evidence for its existence in vivo.

Senescence, an Ultimate Anti-proliferative Barrier . . . and Beyond

Cellular senescence is a state of permanent cell cycle arrest characterized by distinct metabolic activity and dramatic changes in cell morphology.^1,2 Originally, its causes were proposed to be due to physiological erosion of telomeres that occurs upon extended cell proliferation (replicative senescence), as well as inappropriate oncogenic signaling (premature senescence). Senescence can be also induced by genotoxic agents, which like eroded telomeres or hyperactive oncogenes, trigger the DNA damage response (DDR) pathways and activate cell cycle checkpoints to halt proliferation. The ultimate DDR targets are cyclin dependent kinases (Cdk), key regulators of cell cycle progression. In addition to directly controlling the onset and progression of DNA replication and mitosis, Cdks phosphorylate the retinoblastoma (pRb) tumor suppressor and the related “pocket-proteins” p107 and p130, thus maintaining them in an inactive state. Continuous Cdk-mediated inactivation of pRb family proteins is a prerequisite for cell cycle progression, because active (hypo-phosphorylated) pocket-proteins drive cell cycle exit by sequestering E2F family transcription factors and repressing the genes required for DNA replication and cell division.³ Deregulation of pRb and of the associated pathways plays a central role in tumorigenesis.⁴ The action of Cdk is closely intertwined with that of another tumor suppressor, p53 (TP53), which is a major target of the PI3K-related protein kinases (PI3K) ATM and ATR and that plays a pivotal role in DDR⁵ (Fig. 1). Following activation by panoply of stress signals, p53 induces the expression of the Cdk inhibitor p21^Waf1/Cip1 (p21), an essential mediator of senescence (see below). More recently, mTOR (mechanistic target of rapamycin), another member of the PI3K family that controls cell growth in response to nutritional cues, has also been implicated in senescence, thus linking cellular growth with the p53 and DDR pathways.^6-9

Figure 1.

Molecular pathways involved in irreversible G1 arrest proceeding senescence. Dysfunctional telomeres, hyperactive oncogenes and various genotoxic agents stimulate the ATM/ATR-mediated DNA damage response (DDR) pathway. The ATM/ATR kinases induce the early response by activating Chk1/2 checkpoint kinases that, in turn, by inhibiting Cdc25 phosphatases, prevent activation of cyclin E1-Cdk2, which is a key regulator of G1/S transition. ATM also phosphorylates the p53 tumor suppressor that, by inducing the Cdk inhibitor p21^Waf1/Cip1 (p21), plays a central role in the G1 exit program (late response). In addition to cyclin E1-Cdk2 (cE1-K2) complexes, p21 also inhibits cyclin D1-Cdk4/6 (cD1-K4/6) complexes that phosphorylate and inactivate pRb family pocket proteins. In turn, active pRb inhibits the E2F1-dependent expression of genes controlling G1/S progression, thus irreversibly blocking the cell cycle entry. Senescence is also associated with p16^Ink4A (p16) upregulation, but the pathways leading to its induction are not entirely elucidated. This Cdk inhibitor specifically targets Cdk4/6 and prevents their association with D-type cyclins. While p16 does not intervene in G1 arrest (p16 is upregulated after induction of p21 and G1 arrest), it plays a key role in senescence maintenance.

Senescence is increasingly considered to be an essential early line of defense against tumor development by preventing proliferation of cells with damaged DNA,^10-12 Indeed, the markers of activated DDR are found in many pre-cancerous lesions, and mutations compromising DNA damage checkpoint (for example, defects in ATM/p53 pathway) are suggested to be a prerequisite for tumor progression.^13,14 Moreover, senescent cells are observed in pre-cancerous lesions in mice and in humans, but not in the subsequent malignant stages. This is consistent with the hypothesis that senescence restricts tumor progression during the early stages of tumorigenesis, whereas progression to malignancy would involve evading senescence.¹ However, senescence has also been implicated both in promoting carcinogenesis¹⁵ and in age-related pathologies^16,17 through its autocrine and paracrine effects, known as the senescence-associated secretory phenotype (SASP) originally described by the Campisi group.^18,19 SASP factors can act cell autonomously by reinforcing cell-cycle arrest and non-cell autonomously by influencing cells in the surrounding environment and promoting immune surveillance, leading to the elimination of senescent cells.^20,12 While in some cases, SASP factors secreted by pre-malignant senescent cells lead to their clearance by the immune system, resulting in tumor regression,¹⁷ in other cases, pro-inflammatory SASP molecules might support carcinogenesis by promoting epithelial-mesenchyme transition and invasion, tumor vascularization and abnormal cell morphology (reviewed in ref.²). Therefore, senescent cells are not merely non-dividing bystanders but, due to their active metabolism (reviewed in ref⁸), they can elicit pro-and/or anti-tumorigenic responses, depending on the microenvironment.¹² Finally, recent work has shown that senescence contributes also to tissue remodeling during embryonic development,^21,22 thus highlighting its importance in non-pathological situations. In conjunction with its role in preventing fibrosis,^23,24 these observations also raise the question as to whether fundamentally different mechanistic pathways underlie these different physiological roles of senescence.

Senescence as an Irreversible G1 Arrest

Originally, cellular senescence was defined as an irreversible arrest in the G1 phase of the cell cycle (reviewed in refs.^25,26). It is initiated by inactivation of cyclin D1-Cdk4/6 and cyclin E1-Cdk2 complexes that positively regulate G1/S phase progression.^27,28 The best-characterized role of cyclin D1-Cdk4/6 complexes is in triggering pRb phosphorylation,^29,30 while cyclin E1-Cdk2 complexes control the onset of DNA replication by directly phosphorylating a number of replication proteins (reviewed in ref.³¹). Like in proliferating cells exposed to DNA damage,^32,33 the ATM/p53/p21 pathway triggered by telomere erosion or activated oncogenes plays a central role in senescence-associated G1 arrest (Fig. 1). Indeed, overly short telomeres elicit the same response as DNA breaks, leading to activation of the DDR signaling cascade and blocking the G1/S transition.^34-36 On the other hand, inappropriate oncogene activation causes a proliferative burst and rounds of error-prone DNA synthesis. The DNA damage accumulated during such deregulated DNA synthesis also activates p53 via the ATM and ATR damage signaling pathways, leading to p21 induction and permanent cell cycle arrest.^37,38

In addition to directly preventing DNA replication by inhibiting Cdk2, p21 also blocks the inactivating phosphorylation of pRb by Cdk2 and Cdk4/6 (Fig. 1), which is a key mechanism in the senescence program.³⁹ Indeed, absence of pRb phosphorylation,⁴⁰ together with the presence of G1 cyclins (cyclins D1 and E1)²⁷ and the lack of mitotic cyclins (cyclin B1),⁴¹ is an early-recognized hallmark of senescence. Very soon after its discovery, p16^Ink4a (p16), another pRb regulator that specifically inhibits the cyclin D1-associated kinases Cdk4 and Cdk6, was also implicated in senescence^28,42-44 (Fig. 1). Together with β-galactosidase staining,⁴⁵ p16 is increasingly used as a senescence biomarker.^16,46,47 However, despite sharing common targets, the 2 Cdk inhibitors have distinct functions and biological roles. Unlike p21, p16 is not required for cell cycle arrest associated with replicative senescence, and its late induction, which occurs only after p21-mediated cell cycle arrest, was suggested to stabilize rather than induce the senescent state.^28,42,48-50 In agreement with p16's role as a guardian of senescence, mutations, silencing or loss of CDKN2A, the gene encoding p16, facilitate senescence bypass and it is frequently associated with carcinogenesis (reviewed in.^51,52 Although this feature earned p16 the rank of tumor suppressor, the results showing high-level p16 expression in number of cancers (reviewed in ref.⁵²) and in tumor-infiltrating immune cells⁴⁷ suggests a more nuanced view regarding its role in senescence and cancer development. Moreover, while p53 directly controls p21 induction, at least during DDR, the mechanisms controlling p16 up-regulation in senescence seem to be p53-independent and have not been completely elucidated (reviewed in ref.⁵¹)

Senescence in G2 – an Old Concept Awaiting Wider Recognition

The idea that senescence could also be induced during a prolonged G2 arrest emerged after the discovery that p21 is also involved in the G2/M checkpoint regulation.^53-55 According to the prevailing model,⁵⁶ DNA damage-induced G2 arrest is triggered by the ATM/ATR-activated checkpoint kinases Chk1/2 that inhibit Cdc25 phosphatases, which are activators of the key mitotic regulator, cyclin B1-Cdk1 (Fig. 2). Based on data obtained using HCT-116 colorectal cancer cell line, it was initially proposed that p21 could play a role in the “maintenance” of DNA damage-induced G2 arrest. Indeed, after transiently arresting in G2, HCT-116 cells in which p21 was knocked out eventually underwent aberrant mitoses, leading to polyploidy or cell death.^53,55 In this work, however, the mechanism whereby p21 sustains G2 arrest was not elucidated. The uncertainty regarding the role of p21 in the G2/M checkpoint was further supported by the findings that in p53-deficient cells arrested in G2, cyclin B1-Cdk1 complexes are inactive⁵⁷ and that even after its overexpression in U2OS osteosarcoma cancer cells, p21 only weakly associated with cyclin B1-Cdk1).⁵⁸ Based on these observations, p21 was not considered to be a crucial component of the G2 arrest mechanism.⁵⁶

Figure 2.

DNA damage response pathways involved in the G2 exit program leading to senescence. Pro-senescence stimuli activate the ATM/ATR DNA damage-signaling pathway, leading to stable G2 arrest that entails mitotic bypass (see also Fig. 3) and permanent cell cycle arrest in a tetraploid G1 phase. The early response involves Chk1/2-mediated inhibition of Cdc25 phosphatases, which promote mitosis by activating cyclin B1-Cdk1 complexes (cB1/K1). In p53-deficient cells, this pathway transiently blocks G2/M progression, but the p53/p21 pathway is required to stabilize G2 arrest. In addition to inhibiting cyclin B1-Cdk1 complexes, p21 mediates premature activation of APC/C-Cdh1, leading to degradation of cyclin B1 and other mitotic regulators. Like in G1 arrest, p21 also activates pRb by blocking Cdk-mediated pRb phosphorylation. Active pRb inhibits the expression of genes that control G2/M progression, leading to irreversible G2 arrest of the cell cycle. It is proposed that p53 might induce senescence independently of p21⁹³ through mechanisms that are not fully clear. Like in G1 arrest, p16 stabilizes senescence in G2 presumably by targeting pRb kinases. However, the pathways involved have not been elucidated.

A breakthrough in understanding how p21 might exercise its G2 function(s) came from the study of normal human fibroblasts (NHF) exposed to ICRF-193, a topoisomerase II inhibitor that specifically causes G2 arrest. Prolonged exposure to this drug led to p53/p21-dependent irreversible cell cycle exit in G2 that was associated with down-regulation of both cyclin A and cyclin B1 and accumulation of non-phosphorylated pRb family pocket proteins.⁵⁹ G2-arrested fibroblasts, in which either p53 or pRb was compromised by viral oncogenes, failed to permanently exit the cell cycle and expressed mitotic regulators even after prolonged exposure to genotoxic agents. This suggested that p21 induces senescence upon G2 arrest by blocking pRb/p130 phosphorylation. (Fig. 2)Similar results were obtained in carcinoma cell lines that conditionally express p53 or following exposure to γ-irradiation or genotoxic drugs⁶⁰ but the notion of G2 cell cycle exit was not evoked. Moreover, in response to DNA damage, the expression of G2/M-specific genes is reduced in a p53- or pRb-dependent manner.^61,60,62 The discovery of G2/M targets of E2F1⁶³ and of the pocket protein-dependent transcriptional repression of mitotic regulators following stable G2 arrest ⁶⁴ provided additional credibility to the G2 exit paradigm. Based on these findings, it was proposed that the “G2 exit program” might serve as a safeguard mechanism to prevent cell adaptation to the G2 checkpoint and consequently the passage into mitosis of cells with damaged DNA (Fig. 2;⁵⁹). Noteworthy, the “G2 exit program” could in principle apply to replicative senescence because telomeric attritions are expected to become apparent (and to be detected as DNA damage) at the end of the S phase (reviewed in ref.³⁶). Mechanistically, G2 exit is distinct from p53/p21-dependent tetraploid G1 arrest that is caused by adaptation to the spindle checkpoint (mitotic slippage) or by cytokinesis failure and that leads to cells with 2 G1 nuclei^65-67 (reviewed in ref.⁶⁸).

Although clearly essential for senescence,³⁹ blocking pRb phosphorylation is not the only role of p21 in the G2 exit program. In primary fibroblasts exposed to genotoxic drugs, in addition to inhibiting cyclin A-Cdk1/2 complexes,⁵⁹ p21 also associates with and sequesters inactive cyclin B1-Cdk1 complexes in the nucleus, thereby preventing mitotic entry⁶⁹ (Fig. 2). This mechanism was proposed to be paramount for cell cycle exit in G2 because, unlike transient inhibition of Cdk1 activation through the Chk1/2 pathway, p21 completely and irreversibly inhibits cyclin B1-Cdk1 complexes, thus rendering impossible their activation by CAK, PLK1 or Cdc25 phosphatases (Fig. 2). The pre-mitotic nuclear localization of cyclin B1 was rather unexpected because cyclin B1 nuclear entry is invariably associated with mitosis onset, whereas cytoplasmic cyclin B1 is thought to be a hallmark of G2 arrest.⁷⁰ Although also observed after ectopic over-expression of p21 or p53 in tumor cell lines^58,71, cyclin B1 nuclear localization associated with DNA damage-induced G2 arrest was not reported in other studies.^53,55 This is probably because most of these works were carried out in p53-proficient tumor cell lines, such as HCT-116 or U2OS cells, in which p21 induction in G2 is inefficient due to impairment of the ATM pathway.⁷² A notable exception was the finding that p21/p27-dependent nuclear sequestration of cyclin B1-Cdk1 complexes also occurs in serum-deprived mouse embryonic fibroblasts (MEFs) lacking expression of all 3 retinoblastoma protein family members.⁷³ However, in this study, cyclin B1 nuclear localization was associated with reversible, rather than irreversible, cell cycle exit.

Originally, downregulation of mitotic regulators following G2 arrest was attributed to pRb-mediated suppression of their expression (Fig. 2). However, subsequent work in several laboratories provided evidence that this is also due to protein degradation triggered by anaphase-promoting complex/cyclosome (APC/C) and its co-activator Cdh1^74-76 (Fig. 2). This premature and p21-dependent activation of APC/C-Cdh1 in G2-arrested cells was thought to be due to down-regulation of the APC/C inhibitor Emi1,⁷⁴ although an implication of the Cdc14B phosphatase in this process was not formally excluded.⁷⁷

Despite the accumulating evidence, the G2 role of p21 and the concept of senescence in G2 failed to gain wider acceptance and are conspicuously missing in a number of influential reviews.^5,36,68,78

Senescence in G2 Comes of Age – New Insights

Several recent publications have confirmed these old observations and provided additional evidence supporting the existence of a G2 exit program and its role in senescence. First, 2 reports showed that replicative senescence could also be triggered in the G2 phase. The first paper demonstrated that the presence of a significant fraction of senescent fibroblasts with a 4N DNA content and strongly expressing cyclin D1 is not due to gradual accumulation of bi-nucleate or tetraploid G1 cells in aging cultures, but the consequence of cell cycle exit after G2 arrest.⁷⁹ Accumulation of cyclin D1 and E1 was also observed in G2-arrested cells following ectopic induction of p21.⁸⁰ More recently, it has been shown that in telomerase-negative cells eroded telomeres elicit DDR predominantly in G2, leading to p53/p21-dependent cell cycle arrest/exit with 4N DNA content.⁸¹ Unlike functional telomeres that trigger a transient DNA damage response in G2 that is required for their protection,⁸² telomeric DNA damage in aging cells persists at a subset of the shortest telomeres and fully activates DDR.⁸¹ These observations corroborate the idea that telomeric attrition might preferentially activate the G2/M checkpoint.³⁶ However, one might wonder why not all senescent cells are stably arrested in G2. Although several explanations are plausible, the most likely is that the G2/M checkpoint is not as robust as the G1/S checkpoint, which relies on strong p21 induction (reviewed in ref.⁷⁸). As a consequence, some damaged cells can escape the G2/M checkpoint, progress into mitosis and arrest in the subsequent G1 phase by the more efficient G1/S checkpoint.

In agreement with this hypothesis, a recent report showed that, unlike genomic breaks, DNA damage signaling elicited by partial telomere deprotection is insufficient to fully activate the ATM-mediated G2/M checkpoint.⁸³ Instead, after completion of mitosis, DNA damage induces p53-dependent cell cycle exit in G1. Why do intermediate-state telomeres not activate the checkpoint and block G2/M progression? The authors suggested that the culprit is altered DDR and insufficient Chk2 activation. However, it is not clear how DNA damage can fully activate ATM and many of its targets (such as H2AX and 53BP1), but not Chk2. Moreover, the authors argued that their experimental model mimics replicative senescence, which is also associated with G1 cell cycle exit and weak Chk2 phosphorylation (based on the analysis of threonine 68 phosphorylation). A major caveat of this interpretation is that Chk2 does not seem to play an essential role in G2 arrest^84,85,72 and it is more likely that damaged cells progress into mitosis due to inefficient Chk1 activation⁸⁶ and/or lack of robust p21 induction. It should be noted that, at least in non-transformed cells, DNA damage-induced Chk2 phosphorylation at Thr68 is very transient⁷² and hence not the best indicator of Chk2 activation (as opposed to SDS-PAGE shift). Furthermore, it is unclear to what extent this telomere-deprotection-DNA damage response actually contributes to replicative senescence because both Chk1 and Chk2 are activated by short telomeres,^34,87,81 whereas telomere erosion in aging fibroblasts activates mainly the G2/M checkpoint.⁸¹ Another intriguing result is the absence of mitotic bypass and endoreplication in cells that lack p53 and that experience prolonged DNA damage signaling due to persistent loss of telomere protection,⁸³ which is in stark contrast with the results documented by the De Lange team⁸⁸ (see below).

With respect to the expression of cell cycle regulators, the phenotype of cells arrested in G2 is similar to that of cells arrested in G1, notably, absence of G2/M cyclins and accumulation of G1 cyclins. What is molecular basis for that? A model originally proposed by the Lahav group suggests that cells stably blocked in G2 eventually undergo mitotic bypass - a passage from G2 to G1 without chromosome segregation - and enter a tetraploid G1 phase where they cannot replicate due to p21-mediated inhibition of cyclin E1-Cdk2.⁸⁹ The absence of mitotic regulators in these 4N cells could be explained by both their degradation and p53/pRb-mediated repression (Fig. 2). This model has been confirmed experimentally by several recent publications. Ye el al. showed that radiation-induced cellular senescence results from mitotic bypass of stably G2-arrested cells into a tetraploid G1 phase, where they accumulate cyclin D1.⁹⁰ Conversely, in p53/pRb-deficient MEFs, DNA damage generated by persistent telomere dysfunction induces prolonged G2 arrest that leads to genome reduplication and tetraploidy due to the absence of the p53/p21-dependent G1/S checkpoint (Fig. 3;⁸⁸). Tetraploidization is caused by mitotic bypass that leads to DNA synthesis in the ensuing 4N G1 phase, as shown elegantly using FUCCI imaging. This technique uses fluorescent probes engineered to label a G1-specific marker (CDT1) and a G2-specific marker (geminin) to follow cell cycle progression in live cells.⁹¹ Subsequent work by the same team has shown that tetraploidization triggered by permanently damaged telomeres promotes tumorigenic transformation of mouse cells.⁹²

Figure 3.

Mitotic bypass associated with p21 induction triggers senescence in tetraploid G1 cells. In p53-proficient cells, DNA damage-induced G2 arrest leads to p21-dependent nuclear sequestration of cyclin B1-Cdk1 (cB1) complexes, thus locking them in the inactive state (cf. Fig. 2). Subsequent APC/C-Cdh1-mediated degradation of cyclin B1 and other mitotic regulators precedes mitotic bypass (skip), leading to irreversible arrest (senescence) in a tetraploid (4N) G1 phase, characterized by accumulation of G1 cyclins (cyclin D1, cD1). In p53-deficient cells, in the case of persistent telomere dysfunction, DNA damage induces strong and sustained Chk1/2 activation leading to stable G2 arrest that, via mitotic bypass, leads to endoreplication, resulting in tetraploidy.⁸⁸ In the absence of sustained Chk1/2 activation, p53/p21-deficient cells exposed to genotoxic agents arrest only transiently in G2 and then progress into mitosis and eventually die.^59,93 Similarly, in aging p53/pRb-deficient cells, extensive telomere erosion induces crisis with massive cell death and tetraploidization through endoreplication or mitotic failure.⁹²

The most recent evidence implicating the G2 exit program in senescence has come from 3 publications that also took advantage of FUCCI imaging system. The Nakanishi group showed that replicative senescence or exposure to various senescence-causing stimuli induces p53-dependent mitotic bypass (referred to as mitotic skipping), entailing irreversible cell cycle arrest in the 4N G1 state and accumulation of the senescent markers p16 and β-galactosidase.⁹³ Moreover, stable G2 arrest induced by ionizing radiation (IR) was associated with p53/p21-dependent premature APC/C^Cdh1 activation and degradation of mitotic cyclins, consistent with previous observations.^74-76 However, unlike p53-mediated mitotic bypass, APC/C^Cdh1 activation alone does not appear to be sufficient to induce senescence, that additionally requires repression of mitotic regulators by pRb family pocket proteins (Fig. 2). The finding that p16 is dispensable both for mitotic skipping and induction of senescence, but is required for its maintenance, is in full agreement with the models proposed earlier whereby p21 is a major mediator of irreversible cell cycle arrest.^28,49 Importantly, this work also provided evidence for the occurrence of 4N senescent cells in human melanocyte naevi, the first in vivo senescence model,⁹⁴ suggesting that the G2 exit program has a more general role in senescence. The results showing that mitotic bypass requires p53, as its absence leads to aberrant mitosis and cell death, corroborate the observations by the Karlseder team,⁸³ but they apparently contradict the finding that mitotic bypass in p53-deficient cells leads to endoreplication.⁸⁸ The most likely explanation is that the strong and persistent DNA damage signaling induced by POT1a/b double knock-down⁸⁸ causes sustained Chk1/2 activation that prevents Cdk1 activation (and mitosis), resulting in prolonged G2 arrest. This would enable the mitotic bypass and the onset of a new round of replication (Fig. 3). Conversely, transient Chk1/2 activation, induced partial TRF2 depletion ⁸³ or acute DNA damage (by IR),⁹³ in the absence of (efficient) p21 induction can transiently delay, but not prevent mitosis.

In an accompanying paper, also using the FUCCI system and cell lines that express a cyclin B1 allele with a fluorescent tag, the Medema group investigated the role of p21-dependent cyclin B1 nuclear retention in senescence induction in G2-arrested cells upon IR.⁹⁵ They elegantly confirmed and extended previous observations^58,69 and pinpointed this event as the first step in triggering the G2 exit program, resulting in permanent cell cycle withdrawal (Fig. 3). Surprisingly, and somewhat counter-intuitively, they showed that APC/C^Cdh1–mediated cyclin B1 degradation before cell cycle exit is actually dispensable for the latter. This result is in agreement with observations by the Nakanishi group showing that Cdh1 is not required for senescence onset.⁹³ Identical conclusions were reached in the recent paper by the Lindqvist team using a similar experimental approach.⁹⁶ In addition, by using fluorescent cyclin B1 and video-microscopy, they confirmed earlier observations⁷² that p21-dependent sequestration of cyclin B1 does not take place in U2OS cells. In this p53/pRb proficient cell line, p21 induction is impaired due to inefficient ATM response ⁷² and, probably, mutation of the Wip1 phosphatase that interferes with p53 activation.⁹⁷ It is not clear, however, how p21-mediated nuclear translocation of cyclin B1-Cdk1 complexes can induce senescence. Although this is the first detectable event, it is more likely that main role of p21 lies in its capacity to trigger pRb-mediated repression of cell cycle regulators (Fig. 2, for more information on p21 functions (see ref.⁹⁸). Indeed, as mentioned above, p21/p27-mediated cyclin B1 nuclear retention in mouse fibroblasts lacking pocket proteins does not induce senescence.⁷³ However, at the present time, the first crucial event in the irreversible journey to senescence has not been clearly identified yet.

A final comment should be made on the importance of p21 in inducing senescence upon G2 arrest. While the previously discussed works^93,95 show that transient induction of p53 in G2 (but not in G1) is sufficient to induce mitotic bypass and senescence, this is not the case for p21.⁹³ Moreover, p21 depletion prevents senescence less efficiently than p53 depletion, prompting the authors to suggest that p53 might activate the pRb family independently of p21 (Fig. 5G in ref.⁹³). While this hypothesis cannot be entirely dismissed, a more likely explanation is that p16 might compensate for the absence of p21 in blocking pRb phosphorylation by cyclin D1-Cdk/6 (Fig. 1and Fig. 2), thus justifying the proposed role of the p16-pRb pathway in senescence maintenance. In addition to a previous report,⁴⁸ this alternative possibility is also supported by data showing that cell cycle exit in p21-deficient cells is associated with p16 accumulation (Figure S3B in ref.⁹³). It is unclear, however, why in this experimental system, p16 induction was not observed in p53-deficient cells (Fig. 2C in ref.⁹³). Clearly, further studies are required to identify the missing components of the senescence program in G2.

In conclusion, a flurry of papers in the last 2 y has highlighted the relevance of the G2 exit program in senescence. Moreover, some data also suggest that the G2/M and not the G1/S checkpoint might be responsible for the onset of replicative senescence. The finding that the decision to enter quiescence also occurs at late cell cycle stages ^99,100 might further change our understanding of the importance of the G2/M checkpoint in cell cycle exit.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Acknowledgments

The authors wish to thank Drs. Etienne Schwob, Jacques Piette and Daniel Fisher for their encouragements to write this review and D.F. and J.P. for critically reading the manuscript. Finally, we thank Dr. Liliana Krasinska for help with the figures.

Funding

This work is supported by the grant E22013.LNCC/DF to Daniel Fisher's team (V.D.) and SIRIC N°095294 to Pierre Roux's team (V.G.).