Review – at the crossroads of life and death

Shirley Polager and Doron Ginsberg

The Mina and Everard Goodman Faculty of Life Science, Bar Ilan University, Ramat Gan 52900, Israel

The tumor suppressor, pRb, restricts is cell-cycle regulated, with dissociation leading to dere- cell-cycle progression mainly by regulating members pression and activation of E2F-regulated and S- of the E2F-transcription-factor family. The Rb pathway phase entry. is often inactivated in human tumors, resulting in Deregulated-E2F activity occurs in the vast majority of deregulated-E2F activity that promotes proliferation or human tumors through several different mechanisms. cell death, depending on the cellular context. Specifi- These include functional loss of pRB; amplification of cally, the outcome of deregulated-E2F activity is deter- D, which promotes of pRB; loss of , a mined by integration of signals coming from the cellular cyclin-dependent kinase inhibitor that inhibits the phos- DNA and the external environment. Alterations in cell phorylation of pRB; and expression of the human papillo- proliferation and cell-death pathways are key features of mavirus (HPV) oncoprotein E7, which disrupts pRB–E2F transformed cells and, therefore, an understanding of complexes [6]. the variables that determine the outcome of E2F acti- In , the E2F family comprises eight genes vation is pivotal for research and treatment. In (–8), which give rise to nine distinct . These this review, we discuss recent studies that have eluci- include E2F3a and E2F3b, which are generated by the use dated some of the signals affecting E2F activity and that of alternative promoters [5]. All family members contain a have revealed additional E2F targets and functions, DNA-binding domain and E2F1–5 have a transactivation thereby enriching the understanding of this versatile domain that enables activation of expression. A short transcription-factor family. amino-acid stretch that mediates binding of pRB is embedded within the transactivation domain and, thus, Introduction this domain also functions in repression of . Members of the E2F family of transcription factors are E2F1–6 contain, in addition, a dimerization domain that is downstream effectors of the tumor suppressor pRB and are required for their interaction with a member of the dimer- considered to have a pivotal role in controlling cell-cycle ization-partner family (DP1–DP4). This interaction progression. Initially, studies revealed that E2Fs deter- enables them to bind DNA and function as transcriptional mine the timely expression of many genes required for regulators. E2F7 and E2F8 do not heterodimerize with DP- entry into and progression through the of the cell family members and can bind DNA as homodimers or as cycle. However, it has become clear that transcriptional E2F7–E2F8 heterodimers [7,8]. E2F-family members have activation of S-phase-related genes is only one facet of E2F been categorized into subfamilies based on their transcrip- activity; it is now known that E2Fs both transactivate and tional activity, structure and interaction with pRB-family repress gene expression. Furthermore, E2Fs function in a members, p107 and p130. E2F1, and E2F3a, which wide range of biological processes, including DNA replica- interact only with pRB, constitute one subfamily and are tion, , the mitotic checkpoint, DNA-damage check- often referred to as the ‘activator E2Fs’, because they are points, DNA repair, differentiation, development and believed to function mainly in activating gene expression. [1–4]. Considering this extensive list, it is per- However, this classification is probably an over-simplifica- haps not surprising that the best-studied member of the tion because DNA microarray studies show that activation family, E2F1, exhibits both oncogenic and tumor-suppres- of the so-called activator E2Fs leads to repression of almost sive activities. as many genes as they activate. –8 function mainly in E2F transcriptional activity is modulated by multiple repression of gene expression and are often referred to as mechanisms, the best known being interaction with the the ‘repressor E2Fs’. product of the retinoblastoma (Rb) tumor-suppressor gene, A somewhat puzzling feature of activator E2Fs, and in pRB [5]. This association not only inhibits the ability of particular E2F1, is the ability to induce the two seemingly E2F to transactivate but, also, actively represses transcrip- contradictory processes of cell proliferation and apoptosis. tion through the recruitment of various chromatin modi- Recent studies shed some light on a long-lasting question: fiers and remodeling factors to the promoters of E2F- how does E2F1 ‘decide’ whether to induce cell proliferation responsive genes. These co-repressors include dea- or cell death? These studies show that signals from cetylases (HDACs), histone methyltransferases and DNA damaged DNA direct E2F to apoptosis, whereas signal- methyltransferases [4]. Formation of pRB–E2F complexes transduction pathways, such as the phosphatidylinositol 3 kinase (PI3K)– kinase B (Akt) and epidermal-

Corresponding author: Ginsberg, D. ([email protected]). growth-factor (EGFR)–ras pathways, inhibit

528 0962-8924/$ – see front matter ß 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.tcb.2008.08.003 Available online 18 September 2008 Review Trends in Cell Biology Vol.18 No.11

E2F-induced apoptosis. These findings are of clinical proliferation. Moreover, these data imply that non-immor- relevance to the treatment and prognosis of cancer tal cells might respond to ectopic-E2F expression by indu- patients. In addition, recent studies identify novel func- cing senescence (or apoptosis, as discussed later) to inhibit tions of E2F, demonstrating that it regulates signal-trans- tumorigenesis. duction pathways and autophagy, a pathway that has a Different activator E2Fs seem to have partially over- dual role in cell survival and cell death. The relevance of lapping, but distinct, roles in the regulation of cell prolifer- these novel activities of E2F to cancer development is ation. For instance, both E2F1 and are required for currently being studied. In this review, these recent devel- cell-cycle entry, but only E2F3 is required for continued cell opments are discussed, focusing on the activator E2Fs. proliferation [11]. Combined loss of E2F1, E2F2 and E2F3 completely abolishes the ability of cells to progress through E2F and cell proliferation the and proliferate [12]. Cells lacking E2F1, E2F2 In quiescent cells, the activator E2Fs are bound to and and E2F3 exhibit impaired expression of E2F target genes repressed by pRB. Mitogenic stimuli that elevate cyclin-D that are essential for cell-cycle progression [12], thereby levels promote phosphorylation of pRB by –cyclin- indicating that lack of activation of crucial E2F targets dependent kinase (CDK) 4 or cyclin D–CDK6 complexes, leads to cell-cycle defects. However, cells deficient in E2F3 which prevents pRB from binding E2Fs (Figure 1). Con- exhibit derepression of the tumor suppressor Arf [13]. The sequently, genes repressed by pRB–E2F complexes are Arf gene encodes a protein that stabilizes and activates derepressed and the now-free E2Fs also activate gene by negating the effects of the E3 ligase Mdm2, expression. Many of the upregulated genes proteins which otherwise promotes p53 degradation. Therefore, involved in DNA replication and cell-cycle progression, derepression of Arf in cells lacking E2F3 leads to activation such as DNA polymerases, minichromosome maintenance of p53 and cell-cycle defects, in particular a defect in re- complex components (MCMs), and . In line entering the cell cycle from quiescence. These abnormal- with the regulation of many proliferation-related genes by ities of E2F3-deficient cells are rescued by losing Arf [13]. E2F, overexpression of E2F1, E2F2 or E2F3a induces Similarly, a recent study demonstrates that cells lacking quiescent immortalized cells to enter S phase [5]. These E2F1, E2F2 and E2F3 exhibit p53 activation; the respon- findings underlie the well-established role of E2Fs as pro- siveness of these cells to normal growth signals and expres- proliferative factors. However, E2F1 overexpression in sion of E2F target genes can both be restored by p53 primary fibroblasts does not lead to S-phase entry, but loss [14]. E2F transcriptional-repressor complexes are cru- instead promotes senescence [9,10]. This indicates cial downstream targets of ARF (alternative reading that immortalization is required for deregulated E2F1 frame of cyclin-dependent-kinase inhibitor 2A) and p53- to promote ectopic S-phase entry and uncontrolled cell induced proliferative arrest [15]. Based on these data, it is

Figure 1. Upstream signals direct E2F1 to proliferation or apoptosis. Mitogenic signals and DNA damage activate different signaling pathways that determine whether E2F1 activity will induce transcription of proliferative or apoptotic genes. Mitogenic stimuli elevate cyclin-D levels, leading to repression of pRB activity via phosphorylation by cyclin D–CDK4 or cyclin D–CDK6 complexes. Subsequently, E2F1 is free to activate proliferative or apoptotic genes. When mitogenic signals also switch on the PI3K–Akt pathway or, in some situations EGFR–Ras–Raf signaling, the apoptotic activity of E2F1 is inhibited and it induces mainly proliferation. In response to DNA damage, signals direct E2F1 to activate its apoptotic target genes: pRB acetylation specifically releases E2F1, and not other E2Fs, to enable induction of apoptosis; E2F1 is phosphorylated by ATM and Chk2 and also acetylated and these modifications direct it preferentially to its apoptotic target genes. In addition, binding of Jab1, an E2F pro-apoptotic co-factor, enhances the apoptotic activity of E2F1. Conversely, SirT1 is induced after DNA damage and it can inhibit the apoptotic functions of E2F1.

529 Review Trends in Cell Biology Vol.18 No.11

Figure 2. Downstream targets of E2F1 affect proliferation and survival. On the one hand, E2F1 can induce cell-cycle progression through activation of cell-cycle-related genes, including genes regulating the G1–S transition and mitotic genes. It also regulates the p53– axis, leading to formation of repressive E2F complexes that induce cell-cycle arrest. On the other hand, E2F1 induces apoptosis through stabilization and activation of p53 or through direct upregulation of apoptotic genes. Additionally, E2F1 can induce apoptosis by repressing survival genes and pathways involving Bcl-2, Mcl-1 and NF-kB. A newly identified cellular function regulated by E2F1 is autophagy. The outcome of E2F1-induced autophagy and how it relates to apoptosis remains to be determined.

hypothesized that, in addition to activating pivotal cell- E2F and apoptosis cycle-related genes, E2F1, E2F2 and E2F3 regulate cell In light of its pro-proliferative function, it seems paradox- proliferation by controlling the ARF–p53 axis, which, in ical that one member of the E2F family, E2F1, can also turn, induces E2F-mediated transcriptional repression mediate apoptosis. Ectopic expression of E2F1 leads to (Figure 2). Such a functional link between activator E2F apoptosis in tissue-culture cells and transgenic mice and repressor E2F is supported also by Drosophila mela- [18]. Notably, E2F1-mediated apoptosis has a physiologi- nogaster genetic studies demonstrating that many of the cal role and is not just an artificial result of ectopic expres- cell-cycle-progression defects and proliferation defects of sion; mice deficient in E2F1 exhibit an excess of mature T flies lacking the activator E2F, dE2F1, can be suppressed cells owing to a defect in thymocyte apoptosis [19]. Also, by concomitant loss of the repressor E2F, dE2F2 [16]. E2F1 knockout mice develop tumors, in part, owing to suppressed apoptosis [19]. Crossing the – E2F as a bistable switch Other E2Fs, in particular E2F3, have been shown to The concept of the restriction point, formulated in 1974, is induce apoptosis in some experimental systems [20–22]. that there exists a point in G1, usually 2–3 hours before S- However, E2F3-induced apoptosis is E2F1 dependent and phase entry, when commitment occurs and cells no longer is largely attributed to the ability of E2F3 to transactivate require external stimuli to complete the cell cycle. The the E2F1 gene [23]. Therefore, although apoptosis induc- restriction point seems to be fundamental for normal tion might not be unique to E2F1, other E2Fs might not differentiation and tissue homeostasis and is deregulated trigger apoptosis directly. Loss of either E2F1 or E2F3 in practically all human tumors. A recent study demon- inhibits the aberrant apoptosis and proliferation observed strates for the first time a pivotal role for E2F in the in Rb-deficient mice, implicating these E2Fs in pRB-de- restriction point, showing that the Rb–E2F pathway con- pendent apoptosis [5]. However, an analysis of RbÀ/À stitutes a bistable switch that controls this crucial point in chimeras shows that apoptosis of pRB-deficient cells can the cell cycle [17]. This study shows that once there is a be prevented by the presence of neighboring pRB-expres- sufficient mitogenic stimulus, E2F is activated and main- sing cells, thereby indicating that pRB suppresses apop- tains this ‘ON’ state independently of additional stimuli. tosis in a non-cell-autonomous manner [24]. This finding Thus, the Rb–E2F pathway converts graded growth-stimu- raises the possibility that, similarly, E2F1 loss suppresses latory signals to an all-or-none E2F response. The positive- aberrant apoptosis in RbÀ/À mice in a non-cell-autonom- feedback loops that affect E2F activity (Box 1, Figure I) are, ous manner. These data notwithstanding, in mouse-brain most probably, an essential feature of this bistable switch. epithelium, inactivation of Rb induces p53 activation and Additional studies are required to fully understand the role apoptosis, both of which are inhibited by E2F1 deficiency in of specific activator E2Fs and positive-feedback loops in a tissue-autonomous manner [25]. E2F1 induces apoptosis this bistable switch. via p53-dependent and p53-independent pathways.

530 Review Trends in Cell Biology Vol.18 No.11

Box 1. Feedback loops in the Rb–E2F pathway E2F1-induced p53-dependent apoptosis The p53 tumor suppressor is a key player in the cellular The Rb–E2F pathway has a central role in determining cell prolifera- tion and viability and is, therefore, tightly regulated. One facet of this response to various stresses, including DNA damage and tight regulation is the complex web of feedback loops that exists oncogene activation. After such stresses, p53 elicits one of a within this pathway and between E2F family members (Figure I). few possible responses, such as growth arrest or apoptosis. E2F activates expression of cyclin E, which, together with CDK2, In some settings, E2F1 signaling to p53 is mediated by phosphorylates pRB, further releasing E2F from the inhibitory grip Arf, a gene that is transcriptionally activated by E2F and of pRB. Thus, the E2F–cyclin E–pRB interactions constitute a positive-feedback loop. In addition, expression of the genes encodes a protein that stabilizes and activates p53 [26]. encoding E2F1, E2F2 and E2F3a is regulated by E2F, generating Interestingly, Arf expression is not upregulated by E2F another positive-feedback loop. Consequently, expression of these during normal cell-cycle progression. Only deregulated- three E2Fs is cell-cycle regulated, with a peak of expression at late E2F activity leads to Arf expression and p53 activation. G1 and early S phase [5]. E2F1 also activates the DNA-damage- induced kinases ATM and Chk2, which, in turn, phosphorylate E2F1, However, E2F1 has been shown to induce p53-dependent forming a potential third positive feedback loop. apoptosis in Arf-deficient mice and cells, indicating that The Rb–E2F pathway also includes a few negative-feedback loops. there must be additional, ARF-independent, functional One mechanism to repress the activity of activator E2Fs involves links between E2F1 and p53. Indeed, E2F1 can induce , which is itself an E2F-regulated gene. Upon E2F-mediated phosphorylation of p53 on residues that are phosphory- expression of cyclin A, the cyclin A–CDK2 complex binds to the E2F1, E2F2 and E2F3a proteins during S phase, leading to the phosphoryla- lated in response to DNA damage and, in this way, trigger tion of their dimerization partner (DP) and the inhibition of E2F–DP activation of p53 [27–29]. Accordingly, E2F1 has been DNA-binding activity [60]. E2F1 also upregulates the expression of the found to influence the expression and activity of the deacetylase SirT1, which binds to E2F1 and inhibits its activities, DNA-damage-responsive kinases ATM (ataxia telangiec- constituting a second negative-feedback loop. The physiological tasia mutated) and Chk2 (checkpoint kinase 2), which relevance of these interactions is highlighted by the increase in the transcriptional and apoptotic activity of E2F1 following SirT1 phosphorylate and activate p53 [28–30] (Figure 2). Of note, inactivation [61]. A third possible negative-feedback loop involves ARF, ATM and Chk2 all have p53-independent activities Skp2 (S-phase kinase-associated protein 2), an E2F-regulated gene and, therefore, their activation by E2F1 might also have a [62] encoding a ubiquitin E3 ligase that targets E2F to degradation role in E2F1-mediated p53-independent apoptosis. [63]. However, SKP2 and E2F have also been suggested to function in Because p53 is capable of mediating various stress a positive-feedback loop that affects the restriction point and includes cyclin E–CDK2, pRB–E2F, SKP2 and p27, another pivotal target of responses, the question arises, how does E2F1-dependent SKP2-mediated degradation [64]. In addition, E2Fs regulate the activation of p53 result specifically in apoptosis? In expression of Rb, Arf, p21, p27 and p18INK4C, all of which might addition to activating p53 itself, E2F1 up-regulates the function in negative-feedback loops that limit E2F level or transcrip- expression of pro-apoptotic co-factors of p53, such as ASPP tional activity: ARF mediates degradation of activator E2Fs [65], whereas pRB and the CDK inhibitors limit the activity of the activator (apoptosis stimulating protein of p53) 1 and ASPP2, E2Fs. Feedback loops also exist between E2F and microRNAs (Box 2). thereby biasing p53 to activate pro-apoptotic genes and Importantly, E2F1 positively regulates the expression of the induce apoptosis [31–33]. Furthermore, it has been repressor E2Fs, E2F7 and E2F8. One of their key targets for reported that E2F1 can bind p53 and, thus, stimulate its transcriptional repression is the E2F1 gene; combined ablation of apoptotic function in response to DNA damage [34]. Nota- E2F7 and E2F8 results in an increase in E2F1 levels and activity, leading to apoptosis [8]. It is striking that none of the aforementioned bly, this pro-apoptotic interaction between p53 and E2F1 is negative-feedback mechanisms are sufficient to effectively restrict independent of E2F1 transcriptional activity. E2F1-induced apoptosis in E2F7 and eight double-knockout mice. This pRB is functionally inactivated in most human tumors, indicates that the many negative-feedback loops in the Rb–E2F resulting in deregulated and hyperactive E2F. In this pathway are, at least to some extent, non-redundant, perhaps context, E2F1-induced apoptosis is viewed as a fail-safe because they function at particular conditions or at distinct time frames along the cell cycle. Altogether, the feedback loops enable mechanism that suppresses transformation. In support of both the regulation and the fine-tuning of E2F, which is crucial to its this premise, Rb is often inactivated together with com- proper and timely activity. ponents of the Arf–Mdm2–p53 pathway; this is thought to reflect selective pressures to reduce the otherwise apopto- tic consequences of deregulated-E2F1 activity.

E2F1-induced p53-independent apoptosis E2F1 can also induce apoptosis in the absence of p53. E2F1-induced p53-independent apoptosis is attributed mainly to E2F-mediated up-regulation of various pro-apop- totic genes. The list of such death-inducing E2F1-regulated genes is constantly growing and includes genes encoding Apaf1 (apoptotic protease activating factor 1), , Bcl-2 homology 3 (BH3)-only proteins, and the p53 family member [18]. However, E2F1 mutants lacking the transactivation domain can induce apoptosis efficiently [18], indicating that, at least in some settings, E2F1-de- pendent gene derepression has a pivotal role in cell death. In addition, it has been found that E2F1 sensitizes cells to

Figure I. Positive- and negative-feedback loops in the Rb–E2F pathway. apoptosis by inhibiting survival signals, in particular, those mediated by the nuclear factor-

531 Review Trends in Cell Biology Vol.18 No.11 kB (NF-kB) or by Bcl-2 and its family member myeloid cell between E2F1 and these kinases [28–30]. Moreover, like leukemia 1 (Mcl-1) [18] (Figure 2). Thus, E2F1 directs pRB, E2F1 also undergoes specific acetylations after DNA apoptosis via a combination of activation, derepression damage and it has been suggested that these modifications and repression of gene expression. Furthermore, E2F1 bias E2F1 towards pro-apoptotic targets, in particular p73 might trigger apoptosis in a transcription-independent [42]. Taken together, the current data indicate strongly manner [35]. The exact set of events underlying E2F1- that DNA damage induces changes in E2F1 and its inter- induced apoptosis in each particular case is most probably acting proteins that direct E2F1 to induce apoptosis determined by the cellular context. Moreover, in many (Figure 1). Accordingly, some studies find that, in response cases in which E2F1 is activated and induces proliferation, to chemotherapeutic drugs that damage DNA, cells lacking its activity does not lead to cell death, indicating that the E2F1 or carrying an RNAi directed against E2F1 exhibit apoptotic potential of E2F1 is tightly regulated during reduced apoptosis [40,43], although other studies do not normal growth. In fact, it is widely believed that, in many detect such a reduction. It seems that, even in response cases, the apoptotic activity of E2F1 is an anti-tumorigenic to DNA damage, E2F1-induced apoptosis is context function that is switched on in conjunction with aberrant dependent. and deregulated proliferation. A conserved region of E2F1 known as the marked box domain was shown to have unique pro-apoptotic activity Regulation of E2F1 apoptotic activity that distinguishes E2F1 from other E2Fs [44]. Jab1, origin- Although the pro-apoptotic activity of E2F1 is well estab- ally identified as a specificity factor for c-Jun and JunD lished, some studies indicate that E2F1 might also have transcription factors, interacts specifically with E2F1 via pro-survival activities. For example, E2F1À/À keratino- this domain and mediates E2F1-induced apoptosis [45] cytes exhibit enhanced apoptosis in response to UV-B (Figure 1). Thus, Jab1 represents the first, and so far [36]. Also, in the imaginal wing discs of Drosophila, the only, pro-apoptotic co-factor of E2F1 to be identified. dE2F1 promotes apoptosis of intervein cells in response The mechanism by which Jab1 affects E2F1-induced apop- to DNA damage, but protects cells within the dorsal– tosis remains to be determined. The marked box domains ventral boundary from apoptosis [37]. Thus, it seems that of the other activator E2Fs also mediate protein–protein E2F1 can induce or inhibit apoptosis depending on cell interactions that affect the ability of E2F2 and E2F3a to type, developmental stage and apoptotic stimulus. More- transactivate specific target genes [46,47]. It is tempting to over, loss of pRB results in E2F1-dependent apoptosis in speculate that proteins other than Jab1 bind to the marked some tissues and biological settings, but not in others. box domain of E2F1, thereby affecting the subset of pro- Furthermore, when normal cells are stimulated to grow, moters that it binds and regulates. E2F1 activity is required for initial entry into the cell cycle, Another effector of E2F1-induced apoptosis is apoptotic but the ability of E2F1 to induce apoptosis is blocked, as inhibitor-5 (Api5), which was identified in a genetic screen mentioned. These observations indicate that E2F1- in flies as a suppressor of E2F1-induced apoptosis in vivo induced apoptosis is under strict control. In recent years, [48]. Api5 functions downstream of E2F1 and does not the regulation of E2F1-induced apoptosis has been the inhibit E2F1 transcriptional activity. This E2F–Api5 func- focus of intensive study. The variables determining tional interaction is conserved from flies to humans. Inter- whether active E2F1 leads to apoptosis are now starting estingly, Api5 levels are elevated in many human tumors to be identified. and might enable tumor cells to evade E2F1-induced One of the first signals recognized to induce E2F1 apoptosis [48]. apoptotic activity was DNA damage. This induction occurs Recent studies highlight the role of signal-transduction by several molecular mechanisms that affect the pRB– pathways in modulating E2F1-induced apoptosis: the epi- E2F1 interaction, E2F1 stability and/or the binding of dermal-growth-factor receptor (EGFR)–Ras–Raf signaling E2F1 to promoters of specific E2F-regulated genes inhibits E2F1-induced apoptosis in vivo during the devel- (Figure 1). In general, the domain within pRB that med- opment of Drosophila imaginal eye discs [49] (Figure 1). In iates binding to E2Fs encompasses amino acids 379–792 this setting, the Rb–E2F pathway and the EGFR–Ras–Raf and has been named the pRB pocket. In the case of E2F1, pathway most probably converge on Hid, a key regulator of an additional C-terminal domain of pRB also binds the cell death in Drosophila. Hid is transcriptionally upregu- transcription factor [38]. This secondary binding site lated by dE2F1 [37] and is inhibited by a Ras- and MAPK within pRB inhibits specifically E2F1-induced apoptosis (-activated protein kinase)-induced phosphoryl- and binding of E2F1 to this site is lost after DNA damage ation. An interesting question concerns whether this func- [38]. In particular, DNA-damage-induced acetylation of tional link between the EGFR–Ras and Rb–E2F pathways pRB within this domain results exclusively in E2F1 is evolutionarily conserved between flies and mammals. release, without affecting the binding of pRB to other Ectopic expression of mutant forms of Ras, and consequent E2Fs, thereby promoting E2F1-induced apoptosis [39]. activation of Raf, inhibits apoptosis resulting from a triple Furthermore, in response to DNA damage, E2F1, but knockout of all Rb family members in mice [50], indicating not E2F2 or E2F3, is phosphorylated by ATM and Chk2, that at least part of this functional link is indeed conserved. and these stabilize E2F1 [40,41]. It is not However, it seems to be the PI3K–Akt pathway and not clear whether these two kinases cooperate and whether the MAPK pathway that inhibits E2F1-induced apoptosis they regulate E2F1 activity in addition to stabilizing the in tissue-cultured mammalian cells [44]. Akt inhibits protein. E2F1 also functions upstream of ATM and Chk2; E2F1-induced apoptosis indirectly by phosphorylating therefore, DNA damage triggers a positive-feedback loop topoisomerase II-binding protein 1 (TopBP1), which, as

532 Review Trends in Cell Biology Vol.18 No.11 a consequence, binds E2F1 and represses E2F1-induced that E2F modulates the activity of signal-transduction apoptosis [51]. Additionally, a recent study identified a pathways via transcriptional regulation of upstream com- subset of E2F1 target genes required for E2F1-induced ponents of such pathways. This indicates that information apoptosis that are repressed specifically by PI3K–Akt flows also in the unexpected ‘reverse’ direction, from a signaling [52]. In line with this, in breast and ovarian nuclear transcription factor, E2F, to upstream signaling cancer patients, low levels of expression of this subset of pathways [54]. For example, E2F affects positively the E2F1 target genes are associated with poor prognosis, MAPK–p38 and PI3K–Akt signaling pathways through indicating that modulating the balance between E2F1- transcriptional induction of the apoptosis-signal-regulat- induced proliferation and apoptosis affects the survival ing kinase 1 (ASK1) and the adaptor protein Gab2, respect- of patients [52]. Interestingly, E2F1 activates Akt in ively [53,55]. It is likely that such E2F-mediated human cells, indicating the existence of an Akt–E2F1 transcriptional regulation of signaling pathways does feedback loop by which E2F1 might modulate its own not, in isolation, activate these pathways, but serves to apoptotic activity [53]. sensitize cells, enabling them to respond to sub-optimal Taken together, the aforementioned data concerning stimuli. In support of this notion, activation of the Rb–E2F the regulation of E2F1-induced apoptosis indicate that, pathway sensitizes fibroblasts to bFGF (basic fibroblast in normal cells, in the absence of DNA damage and in the ), most probably through E2F-mediated tran- presence of proper external signals for survival and scriptional upregulation of FGF receptor 1 [56], and ectopic growth, the apoptotic potential of E2F1 is suppressed expression of E2F1 sensitizes cells to PDGF (platelet- (Figure 1). Conversely, when normal cells experience derived growth factor), most probably by E2F-mediated DNA damage or lack of growth and survival signals the transcriptional activation of the MEK (MAPK kinase)– apoptotic activity of E2F is unleashed. In transformed ERK (extracellular signal-regulator kinase) pathway cells, E2F is often deregulated and hyperactive owing to [57]. Most experiments demonstrating E2F-signal-trans- defects in the Rb pathway and this might trigger apoptosis. duction crosstalk were performed in cultured cells using However, E2F-induced apoptosis is inhibited in the trans- ectopically expressed E2F. It remains to be determined formed cells either by mutations in apoptotic pathways whether such functional interactions exist in vivo. Cancer downstream to E2F (for example, a mutation in the p53 cells must acquire the capacity to survive and proliferate in pathway is often detected in cells with a defective Rb an initially unfavorable environment with limited avail- pathway) or by alterations in pathways regulating the ability of growth and survival factors. Therefore, if phys- apoptotic activity of E2F (for example, a high PI3K iological, the effects of E2F on signaling pathways (that activity). control cell survival, cell growth and cell proliferation) Of note, the recent discovery that the EGFR–Ras–Raf might be crucial in conferring a selective advantage on and PI3K–Akt pathways inhibit E2F1-induced apoptosis cells that exhibit deregulated-E2F activity. has potentially important implications for future cancer therapy. Pharmaceutical attenuation of signaling path- E2F and autophagy ways that inhibit E2F1-induced apoptosis might lead to Although the role of E2F in apoptosis is well characterized, the selective killing of tumor cells that lack functional pRB little is known regarding its putative participation in other but retain operational apoptotic machinery downstream of cell-death pathways. Autophagy is an evolutionarily con- E2F1. However, the distinct signaling pathways that served vesicular-trafficking process that mediates the regulate E2F1-induced apoptosis in specific human tissues degradation of cytosolic proteins and organelles and is and tumors must first be better characterized. stimulated rapidly in response to various stresses, such as nutrient or growth factor deprivation. Autophagy is E2F – not only proliferation and apoptosis implicated in controlling cell viability; in some settings, Although most literature concerning activator E2Fs functioning as a survival mechanism whereas, in others, relates to roles in the G1–S transition and apoptosis, there promoting cell death. Interestingly, Rb-deficient fetal hep- are a growing number of reports indicating that they affect atocytes exhibit non-apoptotic death with autophagic fea- additional processes, including mitosis, DNA-damage tures, implicating deregulated E2F in autophagy [58]. checkpoints and DNA repair [4]. Additionally, studies in Furthermore, the E2F-regulated BH3-only protein BNip3 animals and cells lacking distinct E2Fs, or distinct E2F was shown to be required for hypoxia-induced autophagic combinations, demonstrate clearly that members of the cell death [58]. Also, activation of E2F1 upregulates the E2F family have a role in developmental processes and in expression of four crucial autophagy genes, LC3, ATG1, the differentiation of various tissues [4,5]. Some of these ATG5 and DRAM, and enhances basal autophagy [59]. effects are most probably consequences of E2F-induced Conversely, reducing endogenous E2F1 expression inhi- proliferation and apoptosis; however, several studies bits DNA-damage-induced autophagy [59]. These recent indicate that E2Fs regulate differentiation directly [4]. studies implicate the Rb–E2F pathway and, in particular E2F1, in the control of autophagy (Figure 2). Thus, it is E2F and conceivable that autophagy and apoptosis are co-regulated Several signaling pathways initiated by mitogenic stimuli at the transcriptional level. Future studies must address in ultimately converge to inactivate the function of the pRB more detail the relationship between E2F1-induced autop- family, which leads to E2F activation. This exemplifies the hagy and E2F1-induced apoptosis. Of note, the nutrient well-documented flow of information from the cell mem- energy sensor AMP kinase a 2 (AMPKa2), which is an brane to the nucleus. However, several recent studies show inducer of autophagy, is an E2F1 target gene required for

533 Review Trends in Cell Biology Vol.18 No.11

Box 2. Activator E2Fs and microRNAs targets (Box 2). Moreover, as novel functions of E2F emerge, such as regulation of autophagy and signaling, Several recent studies demonstrate complex functional interactions future studies should investigate their effect on the overall between activator E2Fs and various microRNAs (miRs). Activator E2Fs, in particular E2F3, upregulate the expression of a cancer- outcome of E2F activity both in culture and in vivo. related cluster of miRs, miR17–92 [66,67]. Interestingly, two of the Ultimately, a better understanding of how the complex miRs in this cluster, miR-17–5p and miR-20a, target E2F1 [66,67] and, network of signals to and from E2F determines the overall to a lesser extent, E2F2 and E2F3 [66]. Thus, miR17–92 and the outcome of E2F activity will pave the way to targeting the activator E2Fs constitute a negative-feedback loop whereby miR17– E2F pathway in cancer therapy. 92 might affect the balance between E2F-induced proliferation and apoptosis. Specifically, the miR-17–92 cluster has been proposed to inhibit apoptosis by preferentially decreasing E2F1 levels [66,67]. Acknowledgements Expression of the miR17–92 cluster is also upregulated by the We apologize to colleagues whose work could not be cited owing to space oncogenic transcription factor c- [68]. Because c-Myc and E2F1 limitations. Work in the laboratory of D.G. is supported by grants from have been shown to activate transcription of each other, this the Israel Science Foundation (ISF), the Israel Cancer Association and establishes an unusually structured network in which c-Myc the Israeli Ministry of Health. simultaneously activates E2F1 transcription and limits its transla- tion, enabling tight control of E2F1 activity. Additionally, as a result References of the structure of this c-Myc–E2F1–miR17–92 network, miR17–92 1 Ren, B. et al. (2002) E2F integrates cell cycle progression with DNA might function as an inhibitor of a hypothetical E2F1–c-Myc positive- repair, replication, and G2/M checkpoints. Genes Dev. 16, 245–256 feedback loop. 2 Cam, H. et al. (2004) A common set of gene regulatory networks links An E2F1–miR negative-feedback loop has been reported recently metabolism and growth inhibition. Mol. Cell 16, 399–411 for the miR-106b-25 cluster: E2F1 upregulates the expression of this 3 Bracken, A.P. et al. (2004) E2F target genes: unraveling the biology. miR cluster in parallel with its host gene, Mcm7, a known Trends Biochem. Sci. 29, 409–417 proliferation-related target of E2F1. Two miRs in this cluster, miR- 4 Dimova, D.K. and Dyson, N.J. (2005) The E2F transcriptional network: 106b and miR-93, both of which share sequence similarity with the old acquaintances with new faces. Oncogene 24, 2810–2826 aforementioned miR-17–5p target E2F1, leading to its reduced 5 DeGregori, J. and Johnson, D.G. (2006) Distinct and overlapping roles translation [69]. Additionally, the miR-106b–25 cluster interferes for E2F family members in transcription, proliferation and apoptosis. with the expression of the CDK inhibitor p21 and the BH3-only Curr. Mol. Med. 6, 739–748 protein Bim; this impairs the TGFb tumor suppressive pathway in 6 Sherr, C.J. and McCormick, F. (2002) The RB and p53 pathways in gastric tumors [69]. A third E2F–miR functional link has been cancer. Cancer Cell 2, 103–112 suggested for miR-34a, a crucial target of the tumor suppressor p53 7 Di Stefano, L. et al. (2003) E2F7, a novel E2F featuring DP- that mediates some of the anti-proliferative and pro-apoptotic independent repression of a subset of E2F-regulated genes. EMBO effects of p53 [70,71]. The relevant targets of miR-34a have not J. 22, 6289–6298 been fully identified; however, miR-34a downregulates the E2F 8 Li, J. et al. (2008) Synergistic function of E2F7 and E2F8 is essential for pathway in human colon-cancer cells [72] and E2F3 has been cell survival and embryonic development. Dev. Cell 14, 62–75 suggested to be an important target of miR-34a [73]. 9 Dimri, G.P. et al. (2000) Regulation of a senescence checkpoint response Overall, the emerging picture regarding E2Fs and miRs is complex by the E2F1 transcription factor and tumor suppressor. Mol. and intriguing, but the current knowledge of this matter is just Cell. Biol. 20, 273–285 beginning to take shape. We look forward to the full identification 10 Lomazzi, M. et al. (2002) Suppression of the p53- or pRB-mediated G1 and elucidation of the functional relationships between E2Fs and checkpoint is required for E2F-induced S-phase entry. Nat. Genet. 31, miRs and a better understanding of the effects of these interactions 190–194 on tumor surveillance and cancer development. 11 Kong, L.J. et al. (2007) Compensation and specificity of function within the E2F family. Oncogene 26, 321–327 its apoptotic activity [52]. It would be interesting to eluci- 12 Wu, L. et al. (2001) The E2F1-3 transcription factors are essential for date the possible role AMPKa2 has in E2F1-induced autop- cellular proliferation. Nature 414, 457–462 13 Aslanian, A. et al. (2004) Repression of the Arf tumor suppressor by E2F3 hagy and to understand whether and how this affects is required for normal cell cycle kinetics. Genes Dev. 18, 1413–1422 E2F1-induced apoptosis. An open question is whether 14 Timmers, C. et al. (2007) E2f1, E2f2, and E2f3 control E2F target E2F-induced autophagy results in cell death or cell survi- expression and cellular proliferation via a p53-dependent negative val. It seems likely that either outcome is possible depend- feedback loop. Mol. Cell. Biol. 27, 65–78 ing on the autophagic stimulus and cellular context. 15 Rowland, B.D. et al. (2002) E2F transcriptional repressor complexes are critical downstream targets of p19ARF/p53-induced proliferative arrest. Cancer Cell 2, 55–65 Concluding remarks 16 Frolov, M.V. et al. (2001) Functional antagonism between E2F family Clearly, E2F function and regulation are highly sophisti- members. Genes Dev. 15, 2146–2160 cated and, although a vast amount of knowledge has been 17 Yao, G. et al. (2008) A bistable Rb-E2F switch underlies the restriction accumulated concerning this crucial effector of cell fate, point. Nat. Cell Biol. 10, 476–482 18 Stanelle, J. and Putzer, B.M. (2006) E2F1-induced apoptosis: turning many challenges lie ahead. Our understanding has killers into therapeutics. Trends Mol. Med. 12, 177–185 advanced concerning which signals, in particular DNA 19 Iaquinta, P.J. and Lees, J.A. (2007) Life and death decisions by the E2F damage and activation of signal-transduction pathways, transcription factors. Curr. Opin. Cell Biol. 19, 649–657 determine whether E2F activity will induce cell prolifer- 20 Paulson, Q.X. et al. (2006) E2F3a stimulates proliferation, p53- independent apoptosis and carcinogenesis in a transgenic mouse ation or death. These features of cellular context probably model. Cell Cycle 5, 184–190 determine whether E2F functions in vivo as an oncogene or 21 Saavedra, H.I. et al. (2002) Specificity of E2F1, E2F2, and E2F3 in a tumor suppressor. Nevertheless, more studies are mediating phenotypes induced by loss of Rb. Cell Growth Differ. 13, required to fully comprehend the overall output of the 215–225 many positive- and negative-feedback loops in the Rb– 22 Ziebold, U. et al. (2001) E2F3 contributes both to the inappropriate proliferation and to the apoptosis arising in Rb mutant embryos. Genes E2F pathway. Also, the integration of microRNAs into Dev. 15, 386–391 the E2F network, both as regulators and as effectors, 23 Lazzerini Denchi, E. and Helin, K. (2005) E2F1 is crucial for E2F- awaits further study and might indicate useful therapeutic dependent apoptosis. EMBO Rep. 6, 661–668

534 Review Trends in Cell Biology Vol.18 No.11

24 Lipinski, M.M. et al. (2001) Cell-autonomous and non-cell-autonomous 48 Morris, E.J. et al. (2006) Functional identification of Api5 as a functions of the Rb tumor suppressor in developing central nervous suppressor of E2F-dependent apoptosis in vivo. PLoS Genet. 2, e196 system. EMBO J. 20, 3402–3413 49 Moon, N.S. et al. (2006) A gradient of receptor 25 Pan, H. et al. (1998) Key roles for E2F1 in signaling p53-dependent signaling determines the sensitivity of rbf1 mutant cells to E2F- apoptosis and in within developing tumors. Mol. Cell 2, dependent apoptosis. Mol. Cell. Biol. 26, 7601–7615 283–292 50 Young, A.P. and Longmore, G.D. (2004) Ras protects Rb family null 26 Vousden, K.H. and Lu, X. (2002) Live or let die: the cell’s response to fibroblasts from cell death: a role for AP-1. J. Biol. Chem. 279, 10931– p53. Nat. Rev. Cancer 2, 594–604 10938 27 Rogoff, H.A. et al. (2002) E2F1 induces phosphorylation of p53 that is 51 Liu, K. et al. (2006) Regulation of TopBP1 oligomerization by Akt/PKB coincident with p53 accumulation and apoptosis. Mol. Cell. Biol. 22, for cell survival. EMBO J. 25, 4795–4807 5308–5318 52 Hallstrom, T.C. et al. (2008) An E2F1-dependent gene expression 28 Berkovich, E. and Ginsberg, D. (2003) ATM is a target for positive program that determines the balance between proliferation and cell regulation by E2F-1. Oncogene 22, 161–167 death. Cancer Cell 13, 11–22 29 Powers, J.T. et al. (2004) E2F1 uses the ATM signaling pathway to 53 Chaussepied, M. and Ginsberg, D. (2004) Transcriptional regulation of induce p53 and Chk2 phosphorylation and apoptosis. Mol. Cancer Res. AKT activation by E2F. Mol. Cell 16, 831–837 2, 203–214 54 Chaussepied, M. and Ginsberg, D. (2005) E2F and signal transduction 30 Rogoff, H.A. et al. (2004) Apoptosis associated with deregulated E2F pathways. Cell Cycle 4, 392–396 activity is dependent on E2F1 and Atm/Nbs1/Chk2. Mol. Cell. Biol. 24, 55 Hershko, T. et al. (2006) E2F1 Modulates p38 MAPK phosphorylation 2968–2977 via transcriptional regulation of ASK1 and Wip1. J. Biol. Chem. 281, 31 Chen, D. et al. (2005) Apoptosis-stimulating protein of p53-2 31309–31316 (ASPP2/53BP2L) is an E2F target gene. Cell Death Differ. 12, 358–368 56 Tashiro, E. et al. (2003) Overexpression of contributes to 32 Fogal, V. et al. (2005) ASPP1 and ASPP2 are new transcriptional malignancy by up-regulation of fibroblast growth factor receptor 1 via targets of E2F. Cell Death Differ. 12, 369–376 the pRB/E2F pathway. Cancer Res. 63, 424–431 33 Hershko, T. et al. (2005) Novel link between E2F and p53: proapoptotic 57 Korotayev, K. et al. (2008) ERK activation is regulated by E2F1 and is cofactors of p53 are transcriptionally upregulated by E2F. Cell Death essential for E2F1-induced S phase entry. Cell. Signal. 20, 1221–1226 Differ. 12, 377–383 58 Tracy, K. et al. (2007) BNIP3 is an RB/E2F target gene required for 34 Hsieh, J.K. et al. (2002) Novel function of the cyclin A binding site of hypoxia-induced autophagy. Mol. Cell. Biol. 27, 6229–6242 E2F in regulating p53- induced apoptosis in response to DNA damage. 59 Polager, S. et al. (2008) E2F1 regulates autophagy and the Mol. Cell. Biol. 22, 78–93 transcription of autophagy genes. Oncogene 27, 4860–4864 35 Bell, L.A. et al. (2006) DNA-binding independent cell death from a 60 Cam, H. and Dynlacht, B.D. (2003) Emerging roles for E2F: Beyond the minimal proapoptotic region of E2F-1. Oncogene 25, 5656–5663 G1/S transition and DNA replication. Cancer Cell 3, 311–316 36 Wikonkal, N.M. et al. (2003) Inactivating E2f1 reverts apoptosis 61 Wang, C. et al. (2006) Interactions between E2F1 and SirT1 regulate resistance and cancer sensitivity in Trp53-deficient mice. Nat. Cell apoptotic response to DNA damage. Nat. Cell Biol. 8, 1025–1031 Biol. 5, 655–660 62 Zhang, L. and Wang, C. (2006) F-box protein Skp2: a novel 37 Moon, N.S. et al. (2005) Drosophila E2F1 has context-specific pro- transcriptional target of E2F. Oncogene 25, 2615–2627 and antiapoptotic properties during development. Dev. Cell 9, 463– 63 Marti, A. et al. (1999) Interaction between ubiquitin-protein ligase 475 SCFSKP2 and E2F-1 underlies the regulation of E2F-1 degradation. 38 Dick, F.A. and Dyson, N. (2003) pRB contains an E2F1-specific binding Nat. Cell Biol. 1, 14–19 domain that allows E2F1-induced apoptosis to be regulated separately 64 Yung, Y. et al. (2007) A Skp2 autoinduction loop and restriction point from other E2F activities. Mol. Cell 12, 639–649 control. J. Cell Biol. 178, 741–747 39 Markham, D. et al. (2006) DNA-damage-responsive acetylation of pRb 65 Martelli, F. et al. (2001) p19ARF targets certain E2F species for regulates binding to E2F-1. EMBO Rep. 7, 192–198 degradation. Proc. Natl. Acad. Sci. U. S. A. 98, 4455–4460 40 Lin, W.C. et al. (2001) Selective induction of E2F1 in response to DNA 66 Sylvestre, Y. et al. (2007) An E2F/miR-20a autoregulatory feedback damage, mediated by ATM- dependent phosphorylation. Genes Dev. 15, loop. J. Biol. Chem. 282, 2135–2143 1833–1844 67 Woods, K. et al. (2007) Direct regulation of an oncogenic micro-RNA 41 Stevens, C. et al. (2003) Chk2 activates E2F-1 in response to DNA cluster by E2F transcription factors. J. Biol. Chem. 282, 2130–2134 damage. Nat. Cell Biol. 5, 401–409 68 O’Donnell, K.A. et al. (2005) c-Myc-regulated microRNAs modulate 42 Pediconi, N. et al. (2003) Differential regulation of E2F1 apoptotic E2F1 expression. Nature 435, 839–843 target genes in response to DNA damage. Nat. Cell Biol. 5, 552–558 69 Petrocca, F. et al. (2008) E2F1-regulated microRNAs impair TGFb- 43 Urist, M. et al. (2004) p73 induction after DNA damage is regulated by dependent cell-cycle arrest and apoptosis in gastric cancer. Cancer Cell checkpoint kinases Chk1 and Chk2. Genes Dev. 18, 3041–3054 13, 272–286 44 Hallstrom, T.C. and Nevins, J.R. (2003) Specificity in the activation 70 Raver-Shapira, N. et al. (2007) Transcriptional activation of miR-34a and control of transcription factor E2F-dependent apoptosis. Proc. contributes to p53-mediated apoptosis. Mol. Cell 26, 731–743 Natl. Acad. Sci. U. S. A. 100, 10848–10853 71 He, L. et al. (2007) A microRNA component of the p53 tumour 45 Hallstrom, T.C. and Nevins, J.R. (2006) Jab1 is a specificity factor for suppressor network. Nature 447, 1130–1134 E2F1-induced apoptosis. Genes Dev. 20, 613–623 72 Tazawa, H. et al. (2007) Tumor-suppressive miR-34a induces 46 Giangrande, P.H. et al. (2003) Identification of E-box factor TFE3 as a senescence-like growth arrest through modulation of the E2F functional partner for the E2F3 transcription factor. Mol. Cell. Biol. 23, pathway in human colon cancer cells. Proc. Natl. Acad. Sci. U. S. A. 3707–3720 104, 15472–15477 47 Schlisio, S. et al. (2002) Interaction of YY1 with E2Fs, mediated by 73 Welch, C. et al. (2007) MicroRNA-34a functions as a potential tumor RYBP, provides a mechanism for specificity of E2F function. EMBO J. suppressor by inducing apoptosis in neuroblastoma cells. Oncogene 26, 21, 5775–5786 5017–5022

535