|Home | About | Journals | Submit | Contact Us | Français|
Expression of the anaphase-promoting complex/cyclosome (APC/C) inhibitor Emi1 is required for the accumulation of APC/C substrates crucial for DNA synthesis and mitotic entry. We show that in vivo Emi1 expression correlates with the proliferative status of the cellular compartment and that cells lacking Emi1 undergo cellular senescence. Emi1 depletion leads to strong decreases in E2F target mRNA and APC/C substrate protein abundances. However, cyclin E mRNA and cyclin E protein levels and associated kinase activities are increased. Cells lacking Emi1 undergo DNA damage, likely explained by replication stress upon deregulated cyclin E- and A-associated kinase activities. Inhibition of ATM kinase prevents induction of senescence, implying that senescence is a consequence of DNA damage. Surprisingly, no senescence or no extensive amount of senescence is evident upon depletion of the Emi1-stabilizing factor Evi5 or Pin1, respectively. Our data suggest that maintenance of a protein stabilization/mRNA expression positive-feedback circuit fueled by Emi1 is required for accurate cell cycle progression, maintenance of DNA integrity, and prevention of cellular senescence.
The timely transcriptional activation and protein stabilization of cell cycle regulators are crucial for irreversible and error-free cell cycle progression. During G1, these events are limited by the retinoblastoma (Rb) family of proteins, which repress E2F-dependent transcription (11), and by the anaphase-promoting complex/cyclosome (APC/C), which drives the ubiquitin-dependent proteolysis of cyclins (39). Protein accumulation at G1/S therefore ultimately requires inactivation of the Rb protein through phosphorylation by cyclin-dependent kinases. APC/C activity is inhibited by Emi1 to permit stabilization of key substrates, including the mitotic cyclins A and B (20). Importantly, Emi1 is itself an E2F target gene, thus bridging transcription and protein stabilization.
Emi1 protein expression persists from G1/S until early mitosis. Its degradation in prometaphase is triggered upon sequential phosphorylation by cyclin B/Cdk1 and Polo-like kinase 1 (Plk1) kinases, thereby generating a recognition motif for the SCFβTrCP E3 ubiquitin ligase (18, 30, 36). A pool of Emi1 remains expressed at the spindle poles beyond prometaphase to organize spindle pole focusing through the END (Emi1/NuMa/dynein) network (1). During G2 and early mitosis, Plk1 and Cdk kinases are active, and during this time, Emi1 stability is ensured through two proposed mechanisms: binding of Evi5 protein to Emi1 (16) and binding of the Pin1 peptidyl-prolyl cis/trans isomerase to Emi1 (5). Both of these mechanisms obstruct the binding of βTrCP to Emi1, thereby protecting Emi1 from precocious degradation.
The cell cycle expression pattern of Emi1 protein in somatic cells already points to cellular functions for Emi1 in G1/S- and M-phase progression. The biological function of Emi1 has been further studied by ectopic expression of a stable form of Emi1, which results in a stabilization of APC/C substrates, prolonged prometaphase, and eventual mitotic catastrophe (30). This proliferative block seen upon Emi1 overexpression is absent in cells lacking p53, allowing for a further increase in genomic instability (26). In addition, loss of Emi1 was shown to result in a decrease in S-phase cells, presumably because of decreased cyclin A accumulation (20). Loss of Emi1 also leads to rereplication as a consequence of decreased levels of cyclin A and geminin APC/C substrates, both inhibitors of replication origin licensing (27). Importantly, a recent Emi1-knockout approach showed that embryos lacking Emi1 do not survive beyond embryonic day 7.5 and manifest defects in mitosis, while polyploid trophoblast giant cells were unaffected (25). Together, these findings highlight a crucial role for regulation of APC/C activity by the Emi1 protein in both G1/S and mitotic cell cycle phases.
Here we studied the pattern of Emi1 expression in mouse tissues and show that Emi1 is specifically expressed in proliferating Ki67-positive compartments of the hair follicle, spermatogonia, and intestinal crypts. Furthermore, a strict correlation exists between Emi1 expression levels and the proliferative status of cultured cells. In addition, we show that although depletion of Emi1 leads to a general decrease in expression of G1/S markers, including cyclin A mRNA and protein levels, this is accompanied by an unexpected increase in cyclin E message, protein, and associated kinase activities. This finding places cyclin E gene transcription in a category separate from other E2F target messages, potentially implying a previously uncharacterized cellular compensatory response. We speculate that this unbalanced G1/S kinase activity unleashes a replication stress response, and we find that DNA damage precedes eventual cellular senescence in Emi1-depleted cells. Importantly, senescence can be prevented by ATM inhibition, and both DNA damage and senescence responses are more prominent than rereplication upon Emi1 depletion. No such senescence is seen upon Evi5 depletion, emphasizing that Emi1, but not necessarily its regulators, links APC/C regulation with DNA damage-induced senescence. Together, our data suggest a crucial in vivo role for Emi1 in E2F target mRNA and protein accumulation, the coordination of replication with mitosis, and prevention of DNA damage-induced cellular senescence.
HeLa, HCT-116, and U2OS cells were from ATCC and were maintained in Dulbecco's modified Eagle's medium (GibcoBRL) according to standardized procedures. hTERT-RPE1 cells were obtained from Clontech and maintained according to the manufacturer's recommendations. C1 human SNF5-inducible malignant rhabdoid tumor cells were a kind gift from P. Verrijzer (Leiden, The Netherlands) and were cultured as described previously (38). Where indicated, cells were synchronized in mitosis by treatment with 100 ng/ml nocodazole (Sigma) for 18 h, collected by shake-off, washed with phosphate-buffered saline (PBS), and replated in nocodazole-free medium. DNA damage was induced by treatment with 1 μM etoposide (Sigma). ATM kinase inhibitor (KU-55933; Calbiochem) was added to a final concentration of 10 μM.
Bacterially produced maltose-binding protein-mouse Emi1 fusion protein was used to raise polyclonal antibodies in rabbits (Josman, LLC). Antibodies were affinity purified against glutathione S-transferase-mouse Emi1 N terminus (amino acids 1 to 219) and verified by using immunoblotting and immunocytochemistry methods (see the supplemental material). Antibodies from two of four rabbits gave comparable and mouse Emi1-specific results. Anti-human Emi1 (20) and anti-human Evi5 (16) antibodies were described previously. For immunoblotting, cell lysates were prepared in RIPB lysis buffer (100 mM NaCl, 50 mM β-glycerophosphate, 5 mM EDTA, 0.1% Triton X-100, 1 mM dithiothreitol, and protease inhibitors), and 5 to 10 μg was analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis. Antibodies used for immunoblotting analysis were p21 (Pharmingen), p27Kip1 (Zymed), p45Skp2 (Zymed), phospho-Rb (Ser807/811; Cell Signaling), FLAG-M2 (Sigma), and Cdh1 (Lab Vision Corp.). Actin (I-19), cyclin A (H-432), cyclin E (HE12), E2F1 (C-20), p16 (H-156), geminin (FL-209), and p53 (DO-1) antibodies were from Santa Cruz Laboratories.
To assess cellular DNA content, cells were washed twice with PBS and fixed in cold 70% ethanol. Cells were then resuspended in PBS containing 50 μg/ml propidium iodide (Sigma), 200 μg/ml RNase A (Calbiochem), and 0.1% glucose and immediately analyzed by flow cytometry on a FACScan cytometer (Becton Dickinson) using CellQuest software.
Small interfering RNA (siRNA) duplexes (Dharmacon) were transfected at 100 nM final concentrations by using Oligofectamine reagent (Invitrogen) according to the manufacturer's instructions. Target sequences were 5′ ACUUGCUGCCAGUUCUUCA 3′ (for human Emi1), 5′ GCAGAAGCCAUUAUGGGUU 3′ (for human Evi5), and 5′ GCAACGATGTGTCTCCCTATT 3′ (for human Cdh1). Controls were transfections with siRNA duplexes targeting green fluorescent protein sequence (5′ GGCTACGTCCAGGAGCGCACC 3′).
Total RNA from cells transfected with siRNAs was prepared using the RNeasy Mini system (QIAGEN) following the provided instructions. Template RNA (10 ng) was analyzed using SuperScript one-step reverse transcriptase PCR (RT-PCR) analysis (Invitrogen) following the manufacturer's recommendations. Briefly, primers were used at 0.2 μM, cDNA synthesis was for 30 min at 50°C, and annealing temperatures were typically 10°C below primer melting temperatures. RT analysis was semiquantitative in that samples were taken every two cycles and results were analyzed at below-saturation signal intensities (typically 28 to 33 cycles). Primers all spanned exon-intron boundaries to prevent signal interference resulting from DNA priming. Forward and reverse primer sequences were as follows: for Emi1, 5′ TGTTCAGAAATCAGCAGCCCAG 3′ and 5′ CAGGTTGCCCGTTGTAAATAGC 3′ (200 nucleotides [nt]); for Evi5, 5′ GAGATGGAAAGACCCACCCAAG 3′ and 5′ TTGTCGTAGTTCAGCCACAGCAGC 3′ (350 nt); for cyclin A, 5′ AGACCCTGCATTTGGCTGTGAA 3′ and 5′ ACAAACTCTGCTACTTCTGG 3′ (150 nt); for glyceraldehyde-3-phosphate dehydrogenase, 5′ TGGAAATCCCATCACCATCT 3′ and 5′ TTCACACCCATGACGAACAT 3′ (200 nt); for Plk1, 5′ CCAGAGGGAGAAGATGTCCA 3′ and 5′ ATAACTCGGTTTCCGTGCAG 3′ (~300 nt); for cyclin E, 5′ GGAGCCAGCCTTGGGACAATAATG 3′ and 5′ TGTCACATACGCAAACTGGTGCAAC 3′ (580 nt); for E2F1, 5′ CATTGCCAAGAAGTCCAAGAACC 3′ and 5′ ATGCTACGAAGGTCCTGACACG 3′ (250 nt); and for p16, 5′ CAGACATCCCCGATTGAAAGAAC 3′ and 5′ CTCACTCCAGAAAACTCCAACACAG 3′ (300 nt).
Cell lysates were prepared in RIPB buffer, and 500 μg of total protein was incubated for 2 h with 2 μg of antibodies to cyclin E (C-19; Santa Cruz), cyclin A (Upstate), or Cdk2 (M2; Santa Cruz). Specific antigen was captured by using protein G- or protein A-Sepharose beads (Sigma) and washed using RIPB and kinase buffer (50 mM NaCl, 20 mM HEPES, pH 7.2, 10 mM MgCl2, 2 mM EDTA, 0.02% Triton X-100). Beads or 4 units of cyclin B/Cdc2 kinase as a positive control (New England Biolabs) was next incubated with 250 μg/ml histone H1 substrate protein (Upstate) in kinase buffer supplemented with 66 μM Na-ATP and 0.25 mCi/ml [γ-32P]ATP (PerkinElmer). Reaction mixtures were incubated at 30°C for 30 min and analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and autoradiography.
Cellular senescence was detected by staining for acidic (pH 6.0) β-galactosidase activity as described previously (14).
Immunohistochemical staining was performed on 0.4-μm paraffin-embedded tissue sections from skin (cheek), intestines, and testes of C57BL/6 mice. The sections were deparaffinized and antigen was retrieved from them by use of citrate (pH 6.0) buffer and microwaving. Endogenous peroxidase and nonspecific binding were blocked using 3% hydrogen peroxide and Power Block (Biogenex), respectively. The chromogen was 3,3-diaminobenzidine (Biogenex), and Mayer's hematoxylin was used as a counterstain. Primary antibodies were affinity-purified rabbit anti-mouse Emi1 and anti-Ki67 antibody (Abcam), and the secondary antibody was Envision Plus (Dako) anti-rabbit antibody-horseradish peroxidase.
Cells growing on coverslips were fixed in 4% paraformaldehyde, permeabilized with 0.2% Triton X-100, and immunostained using 0.4 to 2 μg/ml primary antibody and Alexa Fluor 488 or Cy3 secondary antibodies (Molecular Probes). Primary antibodies used were against PCNA (PC10; Santa Cruz), phospho-histone H2A.X Ser139 (Upstate), phospho-Chk2 Thr68 (Cell Signaling), phospho-Ser15 p53 (Cell Signaling), or replication protein A (RPA; NeoMarkers). DNA was counterstained with Hoechst 33288 dye and image analysis was performed using a Zeiss AX10 microscope and Slidebook 4.1 software.
Cells treated with siRNA as described above were harvested at 22 or 76 h after transfection and total RNA was prepared using the RNeasy Mini system (QIAGEN) incorporating an additional on-column DNase digestion step. The methods for preparation of cRNA and array hybridization were provided by Affymetrix (Santa Clara, CA). cRNA was hybridized to Affymetrix Human Genome U133 Plus oligonucleotide arrays, which were scanned on a GeneChip 3000 scanner. Data analysis was performed using Affymetrix GCOS 1.4 software. Experiments were performed in triplicate, and averages were compared to those of controls by using routines in the R programming language (script available on request).
Somatic cells express the Emi1 protein from G1/S through early mitosis, thereby defining a window of stabilization and accumulation of critical S-phase and M-phase regulators (20). Together with the fact that Emi1 is an E2F target gene, these data suggested that Emi1 protein is expressed in proliferating cells. We therefore investigated the expression of Emi1 protein in mouse tissues. Affinity-purified anti-mouse Emi1 polyclonal antibody was verified to specifically recognize endogenous mouse Emi1 protein (see Fig. S1 in the supplemental material). Consistent with the biochemical role of Emi1 in G1/S progression, immunohistochemistry of mouse tissues confirmed the expression of Emi1 in proliferating Ki67-positive cellular compartments in the mouse hair follicle and skin epidermis, spermatogonia, and intestinal crypts (Fig. (Fig.1A1A).
Conversely, we asked whether Emi1 expression is reduced in nonproliferating cells, either upon induction of quiescence or irreversible cell cycle exit (cellular senescence). We first studied quiescent human diploid telomerase RT-immortalized retinal pigment epithelial (RPE) cells that were induced to resume cycling by serum addition. As expected, serum-starved RPE cells expressed high levels of the p27Kip1 Cdk inhibitor and low levels of G1/S cyclins, including cyclin E and cyclin A (Fig. (Fig.1B).1B). Emi1 protein was undetectable in quiescent RPE cells but present at times coinciding with cyclin A protein accumulation, consistent with the notion that Emi1 is required to stabilize cyclin A through APC/C inhibition. Of note, levels of the Emi1-stabilizing protein Evi5 were only slightly decreased upon quiescence. This shows that, in comparison with Emi1, Evi5 expression was less significantly affected by proliferative status in this system (Fig. (Fig.1B1B).
The effect of expression of Emi1 in cells induced to undergo irreversible cell cycle exit (senescence) was studied in two different systems. The first used IPTG (isopropyl-β-d-thiogalactopyranoside)-inducible expression of the SNF5 chromatin remodeling factor in malignant rhabdoid tumor cells (C1). These cells lack endogenous SNF5 and represent a well-studied example of inducible senescence that is thought to involve transcriptional induction of p16Ink4a (38). Secondly, we used RPE cells induced to undergo a DNA damage checkpoint response by etoposide treatment. In both cases, a reduction of Emi1 protein levels correlated with cell cycle exit (Fig. 1C and D; cell cycle exit upon etoposide treatment is shown in Fig. S2 in the supplemental material), and again, Evi5 protein levels were not significantly affected, on occasion showing even a slight elevation upon DNA damage. The decrease in Emi1 protein levels of dimethyl sulfoxide-treated cells reflects the fact that proliferating RPE cells eventually become density-arrested when reaching confluence. Together, these data represent the first analysis of Emi1 expression in vivo, demonstrating the presence of Emi1 protein in proliferating cell compartments, and show that Emi1 expression is positively correlated with proliferative status by using in vitro cell systems.
We previously showed that cells lacking Emi1 protein show a reduced S-phase fraction as measured by bromodeoxyuridine (BrdU) incorporation and a delayed accumulation of cyclin A in synchronized cells (20). Here, we extended this analysis and detected a significant reduction in the proliferative capacity of RPE cells treated with human Emi1 siRNA (Fig. (Fig.2A).2A). Analysis of the cell cycle profile by propidium iodide (PI) staining and fluorescence-activated cell sorter analysis demonstrated that cells treated with Emi1 siRNA indeed undergo a delayed progression through S and G2 phases compared to control siRNA transfected cells (Fig. (Fig.2B;2B; quantitations are shown in Fig. S3A in the supplemental material). Importantly, this delay was partially rescued by codepletion of Cdh1 protein, suggesting that this is due to the destruction of APC/C substrates upon Emi1 depletion (Fig. (Fig.2B;2B; see also Fig. S3A in the supplemental material). In addition, a modest increase in cells with a larger than G2 DNA content was visible (up to ~6.3%), indicating that a percentage of cells undergoes endoreduplication. Similar results were obtained using an additional Emi1 siRNA targeting sequence (see Fig. S3B in the supplemental material). Of note, more significant endoreduplication was measured upon Emi1 depletion in transformed HeLa and U2OS cells (see Fig. S3C in the supplemental material), highlighting the fact that checkpoint signaling in primary cells and that in transformed cells are likely different.
To better assess DNA replication dynamics, we next measured BrdU incorporation rates by using flow cytometry. Consistent with the PI profiles, an increased percentage of cells were BrdU positive early after Emi1 depletion, implying slower S-phase progression. Emi1-depleted cells, however, exhibited a reduction in the rate of S-phase entry, evidenced by reduced BrdU incorporation in synchronized cells (see Fig. S3D in the supplemental material). The latter explains published data describing reduced BrdU incorporation in asynchronously growing Emi1-depleted cells (20, 27), as these studies applied longer (2-h) BrdU-labeling pulses (as opposed to a 30-min pulse in our present experiment), allowing the detection of cells both entering and progressing through S phase.
Cells lacking Emi1 were large and flattened and contained relatively large nuclei (Fig. (Fig.2A),2A), a cellular morphology reminiscent of cellular senescence. We therefore examined human Emi1 siRNA-treated RPE cells for the characteristic SA-β-gal marker. Approximately 37 to 65% of RPE cells stained positive for SA-β-gal at 4 or 5 days after Emi1 siRNA treatment (Fig. (Fig.2C),2C), a phenotype that was recapitulated using two additional siRNA targeting sequences (see Fig. S4A in the supplemental material). Importantly, senescence induction upon Emi1 depletion required APC/C activity, as it was rescued upon codepletion of Cdh1 (Fig. (Fig.2D),2D), and was also observed in HCT-116 and HeLa cells (see Fig. S4B in the supplemental material). Quantitation of DNA content by using microscopy showed that enlarged senescent nuclei contained a relative integrated DNA intensity similar to that of control cells (see Fig. S4C in the supplemental material), confirming the flow cytometry data shown in Fig. Fig.2B.2B. Together, Emi1 depletion and consequent APC/C activation lead to the induction of senescence both in RPE cells that resemble primary cells and in transformed cells that lack critical tumor suppressor pathways (the p53 pathway in HeLa cells) or contain oncogenic mutations (the Ras pathway in HCT-116 cells).
To further explore the initial biochemical response to Emi1 depletion, we followed protein expression in synchronized control or Emi1 siRNA-treated cells. As shown previously, Emi1 knockdown results in a decreased accumulation of cyclin A when cells are followed from mitosis to G1/S (Fig. (Fig.3A).3A). The decreased cyclin A accumulation correlated with a decreased phosphorylation and therefore a transcriptionally repressive form of the Rb pocket protein. Correspondingly, decreased abundance of the E2F target protein and APC/C substrate Skp2 was measured (4, 49), as well as a decrease in E2F1 protein, itself an E2F target gene product. In addition, we measured a decrease in the APC/C target protein geminin, an inhibitor of replication licensing (Fig. (Fig.3A).3A). Emi1-depleted cells, however, showed a marked increase in cyclin E1 protein levels, an effect that was also seen by using an additional Emi siRNA oligonucleotide and seen up to 48 h or more after release from nocodazole (see Fig. S4D in the supplemental material). Of note, Cdh1 codepletion restored Rb phosphorylation and cyclin A and E protein levels to control levels (see Fig. S4E in the supplemental material), confirming that these biochemical responses to Emi1 depletion are explained by an increase in APC/C activity. No significant induction of the Cdk inhibitor p16Ink4A was seen, while p27Kip1 levels were only slightly increased in this time course. Taken together, Emi1 knockdown results in APC/C activation and a consequent decrease in E2F target protein expression with the exception of an increase in cyclin E1 protein levels.
Cyclin E-directed Cdk2 and GSK3 kinase activities together mediate cyclin E phosphorylation and subsequent ubiquitination in an autoregulated destruction loop (10, 50, 52). It was therefore surprising to find high levels of cyclin E protein without concomitant increases in its inhibitor p27Kip1. To better understand this result, we analyzed the abundance of cell cycle mRNAs by RT-PCR. Again, an overall reduction in E2F target gene transcription, including transcription of cyclin A, E2F1, and Plk1 genes, was measured, while cyclin E gene transcript levels were significantly increased (Fig. (Fig.3B).3B). This result was confirmed by microarray analysis of RNAs isolated at 22 or 76 h after Emi1 siRNA treatment (Fig. (Fig.3C),3C), which also showed decreases in cyclin B and D1 mRNA abundances. A similar increase in cyclin E mRNA levels was seen with the use of an additional Emi1 siRNA oligonucleotide, emphasizing that this is an on-target effect (see Fig. S4G in the supplemental material). Of note, while p16Ink4A induction is often measured upon cellular senescence, we see no increase in p16Ink4A mRNA at 22 h and only a slight induction at 72 h. Thus, Emi1 expression and consequent APC/C inhibition are required for the accumulation of E2F target proteins and transcripts, with cyclin E as a notable exception.
We next asked whether this increase in cyclin E level correlates with an increase in cyclin E/Cdk2 kinase activity. Cell extracts were subjected to cyclin E or Cdk2 immunoprecipitations, and associated kinase activities were assessed in vitro. Cell extracts from Emi1-depleted cells showed a significant increase in both cyclin E- and Cdk2-associated kinase activities, whereas cyclin A-associated activity was reduced (Fig. (Fig.3D).3D). Together, these data imply that G1/S progression and cyclin/Cdk activities are deregulated in cells lacking Emi1.
Recent work has indicated that deregulated or increased cyclin E/Cdk kinase activities are associated with altered replication dynamics and consequent DNA damage induction (2). Stalled replication leads to the formation of single-stranded DNA intermediates that are visualized by recruitment of RPA (56). We measured a significant increase in cells with distinct nuclear RPA foci upon Emi1 depletion (Fig. (Fig.4A),4A), suggesting replication stress. Premature termination of DNA replication can lead to fork collapse and consequent DNA double-strand breaks, eventually triggering robust activation of the DNA damage checkpoint (43). Indeed, staining of Emi1-depleted cells showed nuclear foci containing phosphorylated histone H2AX (γ-H2AX), a marker of DNA damage foci (Fig. (Fig.4B).4B). In addition, a significantly increased percentage of cells contained the active phosphorylated form of the Chk2 checkpoint kinase (Fig. (Fig.4C),4C), implying activation of the upstream ATM/ATR kinases (31, 43). The amount and size of foci stained with either γ-H2AX or phospho-Chk2 per individual nucleus were measurably greater in Emi1 siRNA-treated cells than those in control siRNA-treated cells, and these two DNA damage markers colocalized in all nuclei that contained the highest numbers of foci (Fig. (Fig.4E).4E). Thus, Emi1 depletion elicits a potent DNA damage response.
Active ATM/ATR and Chk kinases initiate a phosphorylation cascade aimed at halting cells and executing DNA repair processes. This includes serine 15 phosphorylation of the p53 protein, resulting in its stabilization and transcriptional activation (41). Indeed, Emi1 siRNA-treated cells contained an increased percentage of cells with nuclear phospho-Ser15-p53 (Fig. (Fig.4D),4D), which also correlated with the extent of γ-H2AX focus formation (Fig. (Fig.4F).4F). Activation of p53 was further evidenced by an increased expression of the p53 target protein p21 as early as 2 days after siRNA treatment (Fig. (Fig.4G).4G). The data described above suggest that loss of Emi1 results in DNA replication stress, likely involving deregulated Cdk activity, followed by a robust activation of DNA damage responses.
Our data show that Emi1 knockdown leads to DNA replication stress and DNA damage, as well as cellular senescence. These findings are in accordance with recent reports showing a correlation between oncogene-induced senescence in early tumors and a DNA hyperreplication response (3, 13). To show unequivocally that senescence is a direct consequence of DNA damage induced upon Emi1 depletion, we sought to inhibit senescence with the recently described ATM kinase inhibitor 2-morpholin-4-yl-6-thianthren-1-yl-pyran-4-one (KU-55933) (19). A pronounced decrease in the percentage of senescent cells was measured when cells were grown for three days in the presence of KU-55933 before SA-β-gal staining (25% versus 75% in dimethyl sulfoxide-treated cells) (Fig. (Fig.5A).5A). In addition, ATM inhibitor-treated cells displayed a change in morphology from round and flattened (senescent-like) to more elongated and spindle shaped (Fig. (Fig.5A).5A). Further substantiating our findings, senescence was also rescued upon ATM depletion using siRNA treatment (data not shown). Interestingly, addition of the ATM inhibitor to cells treated with Emi1 siRNA resulted in increased cell death (Fig. (Fig.5B).5B). This implies cell death as an alternate outcome to senescence and suggests that checkpoint inactivation (and cell cycle reentry) in the context of senescence-triggering stimuli sensitizes cells to undergo mitotic catastrophe. We conclude that cellular senescence upon Emi1 depletion is the result of an ATM-associated DNA damage response.
We previously showed that depletion of the Emi1-stabilizing factor Evi5 leads to precocious Emi1 protein destruction and consequent cell cycle arrest (16). Indeed, Evi5 siRNA treatment in RPE cells also resulted in a decrease in proliferation (Fig. (Fig.6A).6A). Moreover, Evi5 depletion resulted in a reduction of Rb phosphorylation and E2F target protein expression, recapitulating effects seen upon Emi1 depletion (compare Fig. Fig.6B6B and and3A).3A). However, Evi5-depleted cells contained no measurable increase in cyclin E protein levels (Fig. (Fig.6B)6B) and instead showed a general reduction of all measured E2F target mRNAs, including cyclin E mRNA (Fig. (Fig.6C).6C). These results therefore revealed that Emi1 and Evi5 knockdowns differ with regards to their effects on cyclin E protein and message levels.
We speculate that Emi1 knockdown elicits senescence through deregulating cyclin E activity and replication stress-induced DNA damage. Since Evi5-depleted cells do not show an increase in cyclin E protein and mRNA levels, we predicted that Evi5 siRNA treatment would not induce DNA damage or cellular senescence. This was confirmed by lack of DNA damage foci (see Fig. S5 in the supplemental material) and SA-β-gal staining (Fig. (Fig.6D)6D) upon Evi5 depletion, showing that Evi5 and Emi1 siRNA treatment differ in the eventual cellular senescence response. Similarly, depletion of Pin1, which was recently shown to stabilize Emi1 in Xenopus egg extracts (5), also led to a decrease in Emi1 abundance but no increase in cyclin E levels or full-blown senescence (see Fig. S5 in the supplemental material). To establish the long-term effect of Evi5 depletion on Emi1 protein levels, we analyzed protein levels at days 1 to 6 after siRNA transfection. This showed that Emi1 protein levels increase 4 days after Evi5 siRNA treatment, compared with a sustained reduction of Emi1 protein levels in Emi1 siRNA-treated cells (Fig. (Fig.6E6E).
In summary, Evi5 and Emi1 depletions both result in a cell cycle arrest through reducing E2F target protein and gene abundance. But only depletion of Emi1 triggers cellular senescence as a likely consequence of DNA replication stress-associated DNA damage.
The ability of Emi1 to act as an APC/C inhibitor has clearly been shown through biochemical approaches using human Emi1 (34) and Xenopus Emi1 in egg extracts (40). In accordance with the biochemical function of Emi to ensure accumulation of S- and M-phase regulators, Emi1 depletion experiments in human cells (20, 27) and mouse and Xenopus oocytes (29, 47) have shown a crucial role for Emi1 in G1/S and mitotic progression. We extend these findings and show that Emi1 protein expression in vivo correlates with the proliferative status of the cellular compartment. Importantly, we found that accurate DNA replication mechanisms are compromised in the absence of Emi1 through increased cyclin E/Cdk2 activities and decreased cyclin A/Cdk2 activities. This likely explains why cells lacking Emi1 display a robust DNA damage response and consequently undergo cellular senescence.
Our present findings are in agreement with a previous report showing that activation of the APC/C holoenzyme by the human T-lymphotropic virus type 1-encoded Tax protein predisposes cells to senescence (24). The authors suggest that senescence results from a permanent G1 arrest through increased Skp2 degradation and therefore decreased SCFSkp2 ligase activity towards p21Cip1 and p27Kip1 Cdk inhibitors. We also detected a decrease in Skp2 levels and (transient) p27Kip1 stabilization. However, p21Cip1 levels increased only after 2 days of Emi1 knockdown at a time when p53 levels are elevated. In contrast, the increase in DNA damage foci and block of senescence by ATM inhibition rather suggest that senescence upon Emi1 depletion is a consequence of DNA damage. Our data are also consistent with the finding that reduced E2F activity, through depletion of its obligate dimerization partner DP, triggers senescence (28), although these authors do not decipher the molecular pathway. It would therefore be interesting to ask whether deregulated cyclin E/Cdk2 activity and/or DNA damage occur upon DP depletion.
The most noticeable and unexpected response to Emi1 knockdown is the immediate increase in cyclin E protein and mRNA levels, reflected also by an elevation in cyclin E/Cdk2 kinase activity. High levels of cyclin E and, more generally, deregulation of cyclin-dependent kinase activity in G1, frequently occur in human cancer and may contribute to tumorigenesis through defective S-phase progression and/or centrosome duplication (15, 21). Intriguingly, several recent studies have connected replication stress in early cancers with allelic imbalances and DNA damage-induced senescence, thus directly linking senescence with a DNA hyperreplication response (3, 13). Furthermore, one study applied DNA combing, a technique that utilizes pulse-labeling of newly synthesized DNA, to directly demonstrate cyclin E-associated premature termination of replication forks and induction of cellular senescence (3). Our data support these findings and show that downregulation of potential oncogenes, such as the Emi1 oncogene, can indirectly trigger oncogene-induced senescence. Sustained lack of Emi1 eventually leads to p53 activation and slightly increased p16Ink4a mRNA expression. This response is delayed compared to cyclin E upregulation, potentially reflecting a necessity for sustained DNA replication stress to trigger DNA damage responses. Importantly, transformed cells in which p53 or other tumor suppressor pathways are mutated, such as HeLa or HCT-116 cells, also senesce upon Emi1 depletion. This suggests that it is likely the primary response to DNA damage that elicits permanent cell cycle arrest.
In addition to DNA hyperreplication as a consequence of precocious or increased origin firing, replication stress can also be achieved through misregulation of origin licensing (7). Downregulation of geminin together with a reduction in Cdk2 activity results in DNA endoreduplication (33, 54). At least in the case of Cdk2 inhibition, this is associated with activation of DNA damage checkpoints (55). Levels of APC/C substrates, including cyclin A and geminin (27, 32), are decreased in Emi1-depleted cells, explaining the endoreduplication phenotype seen in cells lacking Emi1. However, at least in our hands, the extent of endoreduplication is cell-type dependent and potentially more penetrant in transformed cells. Furthermore, while at least 40% of RPE cells show signs of DNA damage and senescence, only 6 to 10% of the cells undergo endoreduplication. Our data are therefore most consistent with a model in which sustained cyclin E/Cdk2 activity triggers DNA damage, although replication stress upon DNA endoreduplication likely contributes. Indeed, our observation that cells lacking Evi5 or Pin1 do not upregulate cyclin E and fail to undergo senescence further supports the notion that increased cyclin E/Cdk2 activity is a major constituent of DNA damage-induced senescence. The induction of DNA damage upon Cdk2 downregulation (55), however, precludes a direct assessment of cyclin E/Cdk2 activation on senescence through rescue experiments.
Cyclin E protein abundance oscillates during a normal cell cycle through periodic transcription and cell cycle-dependent protein destruction (21). Cyclin E/Cdk2 activity peaks at G1/S when it is derepressed by inactivation of inhibitors such as p27Kip1, and subsequently, it is eliminated by phosphorylation- and ubiquitin-dependent degradation of cyclin E (22, 35, 44). Since the cyclin E gene is an E2F target gene (8, 17, 37), a crucial outstanding question is why the cyclin E gene, but not the cyclin A gene or other E2F target genes, is upregulated upon Emi1 knockdown. This finding is not without precedent: sustained expression of Cdh1 and therefore sustained activation of APC/C holoenzyme also leads to increased cyclin E and DNA overreplication (42). These authors measured an initial decrease in Skp2 levels and increase in the SCFSkp2 substrate p27Kip1 (9, 45, 46), similar to our findings. However, Cdh1 overexpression eventually led to increased levels of E2F1, possibly explained by decreased cyclin A/Cdk2 activity and consequent failure to inactivate E2F/DP complexes at the end of S phase (23). In contrast, E2F1 levels remain low after Emi1 knockdown, at least until well after cyclin E mRNA and protein levels are elevated. Furthermore, activation of E2F1 would not explain the differential effects on cyclin E and cyclin A E2F target gene mRNAs. Interestingly, cyclin E levels are restored to control levels upon Emi1 and Cdh1 codepletion. We therefore speculate that selective activation of cyclin E may be due to degradation of an unidentified APC/C substrate that selectively regulates cyclin E gene transcription, potentially constituting a compensatory response to decreased cyclin A/Cdk2 activity.
It is interesting that depletion of the Emi1-stabilizing factor Evi5 does not lead to senescence, even though Emi1 levels remain low for a substantial number of days. It therefore seems unlikely, yet not impossible, that this difference is due to compensatory feedback mechanisms that drive Emi1 accumulation upon Evi5 loss. Persistent low levels of Emi1 protein, and perhaps even a localized pool of Emi1 protein, may be sufficient to partially inhibit APC/C activity and cyclin E accumulation and thereby prevent senescence upon Evi5 depletion. Another potential explanation is that Evi5 is actually required to maintain senescence-associated cellular homeostasis, consistent with the continued expression of Evi5 in quiescent as well as senescent RPE cells. Indeed, recent reports have implied a role for Evi5 in recycling endosome trafficking through binding Rab11 (12, 51), broadening its role beyond that of a regulator of Emi1 stability. Intriguingly, the Pin1 prolyl isomerase, in addition to stabilizing Emi1 (5), also participates in cyclin E turnover (48, 53). In mouse embryo fibroblasts, loss of Pin1 is associated with increased cyclin E levels and genomic instability (53). Even though we failed to measure increased cyclin E levels upon Pin1 depletion, a detectable percentage of Pin1-depleted cells underwent senescence. This suggests that a more pronounced or long-term depletion of Pin1 may be required to elicit cyclin E stabilization and senescence responses.
In conclusion, we studied the effect of the Evi5/Pin1/Emi1 proliferation axis on underlying E2F target gene expression. We found that both Evi5 and Emi1 are required for the efficient induction of E2F target genes and accumulation of crucial cell cycle progression proteins. Thus, the “stabilization circuit” is required not only downstream but also upstream of the E2F transcriptional mechanisms. The recent finding that Rb and the APC/C interact physically (6) raises the intriguing possibility that lack of Emi1 promotes this interaction and may place this circuit in the vicinity of Rb-controlled genes. Cells lacking Emi1, but not those lacking Evi5, undergo cellular senescence through evoking a DNA damage response as a consequence of DNA replication stress. These data highlight the fact that accurate balancing of E2F activity, and by implication E2F target gene expression, including expression of the Emi1 gene itself, is required to block senescence. Emi1 therefore acts as a central component of a self-amplifying gene expression and protein stabilization circuit that elicits senescence when it is switched off. Our data further suggest that targeting cellular Emi1, but not necessarily its regulators, may constitute an effective means to halt tumor cell proliferation.
We thank Rob Vries and Peter Verrijzer for providing C1 SNF5-inducible cells and Mei Hsieh for generating maltose-binding protein-Emi1 immunogen. We also thank the Genentech Inc. microarray facilities, Zora Modrusan and Cynthia Honchell for help with microarray analysis, Matt Brauer for statistical analysis, Jeff Eastham-Anderson for help with image analysis, and the Genentech Inc. oligonucleotide synthesis core for generation of siRNA oligonucleotides. We are grateful to T. Halazonetis for communicating results prior to publication, members of the Jackson lab for insightful discussions, and Jorge Torres, Adam Eldridge, Tom O'Brien, and Andrea Cochran for critical reading of the manuscript.
This work was supported by a Damon Runyon Cancer Research Foundation fellowship (DRG-1811-04) to E.W.V., NIH grant K08 NS45077 to N.L.L., and NIH grants RO1 GM054811 and RO1 GM063023 to P.K.J.
Published ahead of print on 17 September 2007.
†Supplemental material for this article may be found at http://mcb.asm.org/.