ATM-dependent E2F1 accumulation in the nucleolus is an indicator of ribosomal stress in early response to DNA damage

Abstract

The nucleolus plays a major role in ribosome biogenesis. Most genotoxic agents disrupt nucleolar structure and function, which results in the stabilization/activation of p53, inducing cell cycle arrest or apoptosis. Likewise, transcription factor E2F1 as a DNA damage responsive protein also plays roles in cell cycle arrest, DNA repair, or apoptosis in response to DNA damage through transcriptional response and protein–protein interaction. Furthermore, E2F1 is known to be involved in regulating rRNA transcription. However, how E2F1 displays in coordinating DNA damage and nucleolar stress is unclear. In this study, we demonstrate that ATM-dependent E2F1 accumulation in the nucleolus is a characteristic feature of nucleolar stress in early response to DNA damage. We found that at the early stage of DNA damage, E2F1 accumulation in the nucleolus was an ATM-dependent and a common event in p53-suficient and -deficient cells. Increased nucleolar E2F1 was sequestered by the nucleolar protein p14ARF, which repressed E2F1-dependent rRNA transcription initiation, and was coupled with S phase. Our data indicate that early accumulation of E2F1 in the nucleolus is an indicator for nucleolar stress and a component of ATM pathway, which presumably buffers elevation of E2F1 in the nucleoplasm and coordinates the diversifying mechanisms of E2F1 acts in cell cycle progression and apoptosis in early response to DNA damage.

Introduction

Transcription factor E2F1 plays roles in cell proliferation and apoptosis through regulating a very diverse array of genes.^1,2 Growing evidences suggest that E2F1 may have a role in maintaining genomic integrity through direct participation in the DNA damage response.^2,3 Supporting this notion, E2F1 associates with the MRN (Mre11–Rad50–NBS1) complex involved in DNA end processing through interaction with NBS1 upon DNA damage, and inactivation of E2F1 causes genomic instability.^3,4 Normally, E2F1 is accumulated in late G₁ phase and rapidly degraded in the S/G₂ phase. Phosphorylation of the retinoblastoma (RB) protein leads to release and activation of E2F1 activity, and E2F1 in association with DP1 behaves as a specific transcriptional activator of cellular genes involved in growth and proliferation.⁵ Upon DNA damage, activated ATM (ataxia telangiectasia mutated) and ATM-related (ATR) kinases phosphorylate E2F1 at S31.^6,7 Chk2 can also phosphorylate E2F1 at S364 upon DNA damage.⁸ Phosphorylation leads to the stabilization and activation of E2F1 upon DNA damage.

The nucleolus is a ribosomal-producing manufactory in eukaryotic cells.⁹ Unlike other organelles in cells, the nucleolus is a dynamic architecture without membrane; instead, it is a large aggregate of rRNA genes (rDNA), precursor and mature RNAs, rRNA-processing enzymes, snoRNPs, and ribosomal protein subunits. The nucleolus disassembles at the beginning of mitosis and reassembles during telophase and early G₁ phase.¹⁰ Assembling ribosomal subunits requires the initial transcription of rDNA by RNA Pol I.¹¹ The initial 47S rRNA precursor transcript is subsequently cleaved to form mature 28S, 18S, and 5.8S rRNAs, which are processed and assembled with the 5S rRNA into ribosome subunits. In addition, the nucleolus has additional functions including mitosis regulation, cell cycle progression and proliferation, stress responses, and biogenesis of multiple ribonucleoprotein particles.¹⁰

Impairment of ribosome biogenesis causes nucleolar stress (also known as ribosomal stress).^12,13 Most types of cellular stresses may disrupt nucleolar integrity, suggesting that the nucleolus is a central hub for stress sensors.⁹ Ultra-structural studies show nucleolar compactness and segregation of the granular and dense fibrillar components in the response to adriamycin (ADR)-induced DNA damage.^14,15 Nucleolar transcription is inhibited under DNA damage stress, in which several proteins regulating rRNA transcription or processing are involved.^16-18 The ARF protein (p14ARF in human, p19ARF in mouse) is predominantly a nucleolar protein that regulates ribosome biogenesis by retarding the processes of early 47S–45S and 32S rRNA precursors through interaction with B23.^17,18 Furthermore, p14ARF interacts with the upstream binding factor (UBF), inhibits its recruitment to Pol I initiation complex, thereby repressing rRNA transcription.¹⁹ In addition, p14ARF physically sequesters HDM2 in nucleoli, relieving nucleoplasmic p53 from HDM2-mediated degradation.²⁰ Multiple ribosomal proteins (RPs) may also bind and inhibit HDM2 (Mdm2 in mouse), leading to activation of p53 responsible for cell cycle arrest or apoptosis.^{10,12,13,20-23} Thus, the nucleolus acts as a sensor for cellular stresses in a p53-dependent and -independent manner.

Most genotoxic agents affect nucleolar structure and function.^10,23-25 Induction of DNA breaks leads to a transient repression in Pol I transcription and a temporary cessation in DNA replication.²⁵ The studies of the correlation of DNA damage with the nucleolus have shown that the nucleolus acts as a sensor for cellular stress signals through stabilization of p53 by RP–Mdm2/HDM2 and ARF–Mdm2/HDM2 interactions, which induces cell cycle arrest or apoptosis.^10,12,20-23 Likewise, E2F1 as a DNA damage-responsive protein also plays roles in response to DNA damage through transcriptional response and protein–protein interaction.^1-4 Furthermore, E2F1 is known to be involved in regulating rRNA transcription.²⁶ In addition, selective inhibition of rRNA may downregulate E2F1, hindering proliferation of p53-deficient cancer cells.²⁷ However, how E2F1 displays in coordinating DNA damage and nucleolar stress has yet to be elucidated. In this study we demonstrate that ATM-dependent E2F1 accumulation in the nucleolus is a characteristic feature of nucleolar stress in early response to DNA damage, which presumably plays a role in coordinating the diversifying mechanisms of E2F1 acts in cell cycle progression and apoptosis in response to DNA damage.

Results

E2F1 accumulates in the nucleolus during early DNA damage

To monitor E2F1 expression under genotoxic stress, H1299 (p53-null) cells were exposed to 1 μM ADR, a topoisomerase II inhibitor that induces DNA double-stranded breaks (DSBs).²⁸ Western blotting showed that following γ-H2AX formation, E2F1 but not E2F2 was upregulated at 6 h, reached a peak at 12 h after ADR exposure, and declined but still was higher than the basal level (0 h) at 24 h (Fig. 1A). The E2F1-regulated cyclin E and p14ARF displayed similarly kinetic expression. To determine the distribution of E2F1 in cellular sub-regions upon DNA damage, the cytoplasm, nucleoplasm, and nucleoli were isolated from H1299 unexposed or exposed to ADR for 6 h, followed by western blotting. In unexposed cells, RB protein distributed mainly in the cytoplasm, but a small amount in the nucleoplasm, while a large amount of E2F1 resided in the nucleoplasm, and an appropriate amount in the cytoplasm and nucleolus. The majority of Nucleophosmin/B23 as well as p14ARF resided in the nucleolus, and Lamin B mainly in the nucleoplasm (Fig. 1B). In exposed cells, RB was significantly decreased in the cytoplasm and nucleoplasm, but E2F1 was increased in the 3 fractions, particularly in the nucleolus, in which p14ARF was also markedly increased. The increased E2F1 accumulation in the nucleolus upon DNA damage was further observed by immunocytochemistry. As shown in Figure 1C and D, more than 70% cells increased nucleolar accumulation of E2F1 at 6 h after ADR exposure, while only about 7% of unexposed cells had increased nucleolar E2F1. The increased nucleolar E2F1 without RB co-localization was also observed at 12 h after ADR exposure (Fig. 1E). This was similar to the positive control experiments, in which exposure of cells to the transcription inhibitor actinomycin D (Act D) for 6 h caused approximately 80% cells with nucleolar E2F1 (Fig. 1D and 1E). The nucleolar E2F1 in most cells was disappeared at 24 h after ADR exposure. These data indicate that E2F1 accumulates in the nucleolus during early DNA damage.

Figure 1. E2F1 accumulates in the nucleolus in early response to DNA damage. H1299 cells were exposed to ADR (1 μM) or Act D (10 nM) for given hours. (A) Western blotting for analyzing γ-H2AX, E2F1, and its targets Cyclin E and p14ARF. After harvest, cells were lysed, and proteins were quantified. Thirty micrograms of proteins were loaded for each lane. The blot was probed with specific antibodies, and the antigen–antibody complexes were visualized by chemiluminescence. β-Actin was used as loading control. (B) Western blotting for analyzing E2F1, RB, nuclear protein Lamin B, and nucleolar protein B23 in cellular sub-regions. The cytoplasm (Cyt), nucleoplasm (Np), and nucleoli (No) were isolated from H1299 unexposed or exposed to ADR for 6 h. (C) Immunofluorescence analysis of the localization of E2F1 in the nucleolus upon DNA damage. After exposed to ADR or Act D (positive control) for 6 h, H1299 cells were fixed and permeabilized. Nuclei were stained with Hoechst 33342 (blue); E2F1 was labeled with anti-E2F1 antibody and FITC-conjugated IgG (green). Cell images were obtained by laser scanning confocal microscopy. Scale bar, 10 μm. (D). Histograms showing the percentage of cells with nucleolar E2F1 in (C) experiments. A minimum of ~200–300 cells was counted for examining nucleolar accumulation of E2F1. Data represent mean ± SD derived from 3 independent experiments. (E) Immunofluorescence analysis of the localization of E2F1 (green) and RB (red) in H1299 at 6, 12, and 24 h after ADR exposure. Scale bar, 10 μm.

The nucleolar accumulation of E2F1 was also evidenced by immunofluorescence labeling of purified nucleoli from H1299 exposed to ADR. The nucleoli from unexposed cells contained an appropriate amount of E2F1 surrounded by B23, but the nucleolar E2F1 staining was markedly increased after exposure (Fig. S1). The early accumulation of E2F1 in nucleoli was also detected in A549 (p53-wt), 2BS, and HCT116-p53^+/+ and -p53^−/− cells exposed to ADR, etoposide (VP16), or 8-Cl-adenosine (data not shown), indicating that the early accumulation of E2F1 in the nucleolus is a common event under genotoxic stresses.

Early nucleolar accumulation of E2F1 is ATM-dependent

We next investigated the effects of the inhibitors of ATM/ATR kinases on the early formation of nucleolar E2F1 after DNA damage. H1299 cells were pretreated with caffeine (100 μM) for 30 min, followed by exposure to ADR for 6 h. Following ATM S1981-phosphorylation activation, E2F1 was upregulated (Fig. 2A) and markedly increased its accumulation in the nucleolus after ADR as well as Act D exposure (Fig. 2B and C). In the presence of caffeine, however, induction of E2F1 by ADR was inhibited, and ADR-induced nucleolar E2F1 was completely abrogated (Fig. 2A–C). To exclude the effect of ATR, we used a lower dose of ATM-specific inhibitor KU55933 (250 nM), which was much less than the dose (over 100 μM) required for inhibiting ATR,²⁹ to pre-treat cells for 6 h. Like caffeine, 250 nM KU55933 significantly blocked ADR-induced E2F1 (Fig. 2D) and abrogated nucleolar accumulation of E2F1 in ADR-exposed cells (Fig. 2E and F) compared with KU55933 untreated cells. These data indicate that ATM kinase activity is required for translocation of E2F1 from the nucleoplasm to the nucleolus.

Figure 2. KU55933 and caffeine abrogate ADR-induced early accumulation of E2F1 in the nucleolus. H1299 cells were pretreated with ATM/ATR inhibitor caffeine (100 μM) or ATM-specific inhibitor KU55933 (250 nM) for 30 min followed by exposure to ADR (1 μM) for 6 h. (A) Western blotting for detecting the effects of caffeine on the auto-phosphorylation/activation of ATM and the induction of E2F1. α-Tubulin was used as loading control. (B) The effects of caffeine on the early nucleolar accumulation of E2F1 in immunofluorescence analysis. Cell images were obtained by laser scanning confocal microscopy. Exposure to Act D as a positive control. Scale bar, 10 μm. (C) Histograms showing the early accumulation of E2F1 in the nucleolus in the presence of ADR and caffeine or Act D in (B) experiments. Data represent mean ± SD derived from 3 independent experiments. (D) Western blotting for detecting the effects of KU55933 on the auto-phosphorylation of ATM and the induction of E2F1. (E) Immunofluorescence analysis of the effects of KU55933 on the early nucleolar accumulation of E2F1. Scale bar, 10 μm. (F) Histograms showing the inhibition of early accumulation of E2F1 in the nucleolus by KU55933 in (E) experiments. **P < 0.01.

To demonstrate further the role of ATM activation in E2F1 accumulation in the nucleolus, we examined ADR-induced nucleolar E2F1 in wild-type primary mouse embryo fibroblasts (MEFs-Atm^wt) and Atm-mutant MEFs (MEFs-Atm^mut), where the mutant Atm could not be autophosphorylated, and E2F1 was not induced in response to ADR exposure (Fig. 3A). Compared with MEF-Atm^wt cells, the nucleolar E2F1 was not detectable in MEF-Atm^mut cells exposed to ADR for 6 h (Fig. 3B–D). Again, these data indicate that the accumulation of E2F1 in the nucleolus during early DNA damage is ATM-dependent.

Figure 3. Nucleolar E2F1 forms in Atm wild-type but not Atm-mutant MEFs. The Atm wild-type primary mouse embryo fibroblasts (MEFs-Atm^wt) and Atm-mutant MEFs (MEFs-Atm^mut) were exposed to ADR (1 μM) for given hours. (A) Western blotting for analyzing ATM, S1981-phospho-ATM, RB, and E2F1. Western blotting was performed as described in Figure 1A. α-Tubulin was used as loading control. (B) Immunofluorescence analysis of the localization of E2F1 in cells upon DNA damage. MEFs were exposed to ADR for 6 h. Cells were fixed and permeabilized. Nuclei were stained with Hoechst 33342 (blue); E2F1 was labeled with anti-E2F1 antibody and FITC-conjugated IgG (green). Scale bar, 10 μm. (C) Histograms showing the percentage of MEFs cells with nucleolar E2F1 in (B) experiments. Data represent mean ± SD derived from 3 independent experiments. (D) Western blotting for analyzing E2F1, Lamin B, and B23 in cellular sub-regions. The cytoplasm (Cyt), nucleoplasm (Np), and nucleoli (No) were isolated from MEFs unexposed or exposed to ADR for 6 h.

Other kinase/enzyme signals do not affect nucleolar E2F1

Next, we asked whether other kinase/enzyme signals that influence E2F1 stability and functions were linked to the formation of nucleolar E2F1. Other groups and our lab have shown that the chemical Gö6976, which can inhibit both CHK1 and CHK2³⁰ or CHK1 only,³¹ is very effective to sensitize cells to DNA damage through abrogating S and/or G₂ arrest. Immunocytochemistry showed that Gö6976 could not abrogate the formation of nucleolar E2F1 in ADR-exposed H1299 at all (Fig. 4A and B), indicating that CHK1 and CHK2 kinase activities are not required to generate nucleolar E2F1. To exclude further the effect of CHK2, we knocked down CHK2 expression in H1299 (Fig. 4C) and treated cells with ADR for 6 h. Compared with control (si-NC transfection plus ADR exposure) cells, nucleolar E2F1 could still be seen in CHK2 knocked-down cells (Fig. 4D). Together, the early formation of nucleolar E2F1 is CHK1/2-independent.

Figure 4. CHK1 and CHK2 kinase activities are not required to generate nucleolar E2F1. (A) The effect of CHK1 and CHK2 inhibitor Gö6976 on the early nucleolar accumulation of E2F1. H1299 cells were pretreated with Gö6976 (250 nM) for 30 min, followed by exposure to ADR (1 μM) for 6 h. After being fixed and permeabilized, the nuclei were stained with Hoechst 33342, E2F1 was labeled with anti-E2F1 antibody and FITC-conjugated IgG. Scale bar, 10 μm. (B) Histograms showing the accumulation of E2F1 in the nucleolus in (A) experiments. Data represent mean ± SD (n = 3). **P < 0.01. (C) Western blotting for detecting knockdown of CHK2 in H1299 cells. Cells were transfected with CHK2 siRNA (siCHK2) (unrelated sequences as control, si-NC). After 24 h transfection, western blotting was performed using anti-CHK2 antibody. (D) The effect of knockdown of CHK2 on the early accumulation of E2F1 in the nucleolus in immunofluorescence analysis.

We also examined the effects of ERK1/2 kinase inhibitor PD98059, JNK inhibitor SP600125, Wnt analog lithium chloride (LiCl), and HDAC inhibitor trichostatin A (TSA) on the formation of nucleolar E2F1 in response to DNA damage. All tested reagents could not abrogate ADR-induced nucleolar E2F1 (Fig. S2), indicating that these kinase/enzyme signals are not required for localizing E2F1 in the nucleolus in early response to DNA damage.

Mutant of E2F1S31A abrogates formation of nucleolar E2F1

Since phosphorylation of E2F1 at S31 by ATM stabilizes E2F1 upon DNA damage,^6,7 we thus speculated that S31-phosphorylation of E2F1 was critical to form nucleolar E2F1 upon DNA damage. Because of no commercial antibody against S31-phospho-E2F1 available, we mutated Ser31 to Ala31 and constructed pEGFP-C3-E2F1-S31A and pEGFP-C3-E2F1 plasmids. Western blotting and confocal microscopy showed that both constructs expressed fusion proteins in transfected H1299 cells, in which endogenous E2F1 and p14ARF were also detectable (Fig. 5A), and the fusion proteins could relocated from the nucleoplasm to the nucleolus in the presence of ADR (Fig. 5B). The numbers of cells with nucleolar EGFP-E2F1 were increased from approximately 41% (control) to 75% when EGFP-C3-E2F1-expressing cells were exposed to ADR for 6 h; in EGFP-E2F1-S31A-expressing cells, however, no significant changes in the numbers of cells with nucleolar EGFP-E2F1-S31A were detected before and after ADR exposure (Fig. 5C). Increased distribution of EGFP-E2F1 but not EGFP-E2F1-S31A in nucleoli was also evidenced by western blotting using fraction proteins (Fig. 5D). These data indicate that S31-phosphorylation of E2F1 by ATM is critical to localizing E2F1 in the nucleolus upon DNA damage.

Figure 5. Mutation of E2F1S31A abrogates formation of nucleolar E2F1. (A) Western blotting for expression of EGFP-E2F1 and EGFP-E2F1-S31A fusion proteins in H1299. Cells were transfected with pEGFP-C3-E2F1 and pEGFP-C3-E2F1-S31A plasmids. After transfection for 24 h, cells were harvested for western blotting with anti-GFP (to detect fusion proteins), anti-E2F1 (to detect endogenous E2F1), and anti-p14ARF antibodies. α-Tubulin was used as loading control. (B) Immunofluorescence analysis of the localization of E2F1 in the cells upon DNA damage. H1299 was transfected with pEGFP-C3-E2F1 or pEGFP-C3-E2F1-S31A plasmid, and after transfection 24-h cells were exposed to ADR for 6 h. Scale bar, 10 μm. (C). Histograms showing the accumulation of E2F1 in the nucleolus in (B) experiments. “0” indicates the ratio of cells without nucleolar E2F1; “1” indicates the ratio of cells with nucleolar E2F1. Data represent mean ± SD (n = 3). (D) Western blotting for analyzing the distribution of EGFP-E2F1 and EGFP-E2F1-S31A fusion proteins as well as B23 and Lamin B in H1299 cells.

Early nucleolar accumulation of E2F1 indicates nucleolar stress

DNA damage inhibits nucleolar transcription.²⁵ To investigate the inhibition of rRNA transcription in ADR-induced DNA damage, we performed northern blot using ITS-1 and ITS-2 as hybridization probes (Fig. 6A). Compared with control (Fig. 6B, lane 1), the formation of primary 47S rRNA transcript and 45S intermediate was almost completely blocked by ADR exposure for 6 h (lane 2), which was similar to Act D exposure (lane 3). 41S and 36S intermediates were also markedly reduced by ADR, whereas the 32S and 12S intermediates were not affected, indicating that ADR inhibits the initiation of rRNA transcription but does not influence rRNA elongation and processing. Notably, the ATM inhibitors caffeine and KU55933 could abrogate the inhibitory effect of ADR on primary 47S rRNA transcript (Fig. 6B, lanes 4–7), indicating that the ATM kinase activity is implicated in the inhibition of RNA Pol I transcription in response to DNA damage. It seemed that the inhibitors might inhibit late processing by unknown reason(s), since 32S and 12S intermediates were reduced in the presence of inhibitors. The inhibition of Pol I transcription initiation by ADR was further demonstrated by metabolic labeling of nascent rRNA. The time course of pulse labeling showed that during the first 2 h of incubation in ³²P-orthophosphate, only a little newly synthesized 45S and 32S intermediates were detectable in ADR-exposed cells compared with unexposed cells. Even after ~3–4 h incubation, newly synthesized 45S and 32S intermediates and mature 28S rRNA in H1299 were markedly reduced; in particular, mature 18S rRNA was failed to be detected, and 5.8S rRNA a little within 6 h after ADR exposure (Fig. 6C). This is probably due to the “site-effect” of their encoding genes in rDNA, in which 18S rRNA locates immediately downstream of the promoter (Fig. 6A), and ³²P-orthophosphate could incorporate only into ongoing transcripts when initiation was inhibited by ADR, so that 18S rRNA was rarely incorporated by ³²P-orthophosphate, while 28S rRNA continues to be elongated and labeled. Again, these data indicate that ADR blocks rRNA transcription initiation. Most recently, the fluorescent dye Pyronin Y (PY) was used to identify specific RNA subspecies in cells.³² We used the dye to determine directly the association of rRNA with the nucleolus. The control cells showed E2F1 staining in both nucleoplasm and nucleoli and rRNAs in nucleoli. In ADR-exposed cells, however, PY staining of nucleolar rRNAs was markedly faded, while staining of nucleolar E2F1 was increased (Fig. 6D), indicating that increased accumulation of E2F1 in nucleoli is associated with the inhibition of rRNA synthesis. Alternatively, increased nucleolar E2F1 indicates nucleolar stress.

Figure 6. Increased nucleolar E2F1 indicates ribosomal stress. (A) The organization of the rRNA genes and rRNA transcription and processing in mammalian cells. Human rDNA repeat unit consists of an intergenic spacer that contains the transcription regulatory elements (P, promoter; UCE, upstream control element; Core, core element), sequences encoding the precursor rRNA (47S) and several transcription-terminators (T) at 3′ end of the rRNA genes, and immediately upstream of the rRNA gene transcription start site (+1). The rRNA genes are present in a single transcription unit, transcribed by pol I to yield a 47S precursor rRNA that produces the mature 18S, 5.8S, and 28S rRNAs. ETS, external transcribed spacer; ITS, internal transcribed spacer. (B) Inhibition of rRNA transcription initiation by ADR. In the presence or absence of caffeine or KU55933 pretreatment (30 min), H1299 cells were exposed to ADR (1 μM) or Act D (10 nM) for 6 h. RNA was extracted and subjected to northern blot hybridization, using ITS-1 and ITS-2 as probes. 28S and 18S rRNA stained by methylene blue as loading control. (C) Inhibition of rRNA transcription initiation by ADR in metabolic labeling of nascent rRNA. After ADR exposure for 6 h, H1299 was incubated in phosphate-free DMEM/10% FBS for 30 min and then incubated for ~0.5–4 h in the presence 10 mCi/ml ³²P-orthophosphate, as described in the “Materials and Methods” section. Total RNA was isolated and 1.5 μg total RNA was separated on a 1% agarose formaldehyde gel. Gel was dried, and metabolically labeled RNA was visualized by autoradiography. (D) Inhibition of rRNA transcription by ADR in Pyronin Y (PY)-stained cells. H1299 was exposed to ADR for 6 h followed by triplet staining of nuclei (by Hoechst 33342 in blue), E2F1 (green) and rRNA (by PY in red). (E) Repression of rRNA promoter activation by ADR. H1299 was transfected with pIRES-RP-Luc reporter vector, and 24 h after the cells were exposed to ADR or Act D for 6 h, followed by reporter enzyme assay. The enzyme activity in pIRES-Luc vector transfected cells is normalized to “1”. Data represent mean ± SD (n = 3). (F) Blockage of E2F1 binding to the rRNA promoter by ADR. H1299 cells were exposed to ADR or Act D for 6 h, followed by ChIP assay. The rDNA promoter sequence extending from −126 to +20, which contains the proximal E2F1 site (TCGCGC) at position −103/−98, was amplified by PCR. IgG precipitation of chromatin as control. Data represent mean ± SD (n = 3). *P < 0.05; **P < 0.01.

To reveal the correlation of nucleolar E2F1 accumulation with rRNA transcription, we detected the transactivation of rRNA promoter by E2F1. Transfection of pIRES–RP–Luc reporter vector, in which the transcription of the reporter luciferase gene was driven by the human rRNA promoter, and reporter enzyme assays showed that ADR as well as Act D exposure markedly blocked the activation of the rRNA promoter (Con vs. ADR, P = 0.018; Con vs. Act D, P = 0.032) (Fig. 6E). To demonstrate whether ADR exposure blocked E2F1-dependent activation of rRNA promoter, chromatin immunoprecipitation (ChIP) was performed and showed a significant decrease in binding of E2F1 to the region from −126 to +20 of the rDNA promoter, which contains the proximal E2F1 site that is necessary to the activation of rRNA promoter,²⁶ in H1299 cells after ADR and Act D exposure (Con vs. ADR, P = 0.0034) (Fig. 6F). ADR-reduced binding of E2F1 to the site (TCGCGC) at −103/−98 of the promoter was also demonstrated by EMSA (Fig. S3, lane 3 vs. lane 2). These data indicate that increased nucleolar E2F1 is concomitant with the repression of E2F1-dependent activation of rRNA promoter during ADR-induced DNA damage.

p14ARF–E2F1 interaction blocks E2F1 binding to the rRNA promoter

Interaction of p14ARF with B23 or UBF represses rRNA transcription.^17-19 A previous report³³ and our recent work³⁴ show p14ARF interacting with E2F1 to inhibit its transcriptional activity. We thus speculated that p14ARF might also suppress E2F1-dependent rRNA transcription in response to ADR exposure. As shown in Figure 1A, E2F1 and p14ARF was expressed at moderate levels in control (0 h) cells, but increased within 6–12 h after ADR exposure. Immunocytochemistry showed that although both proteins were detected in the nucleolus of control cells, the nucleolar E2F1 was seemingly in the region of dense fibrillar component (DFC), where rRNA transcription takes place, while p14ARF was essentially limited to the nucleolar periphery matrix (Fig. 7A, right upper panel). In ADR-exposed H1299, however, E2F1 and p14ARF both were increased in the nucleolus, in which E2F1 gathered together to form conspicuous clusters, while p14ARF expanded from nucleolar periphery to the nucleolar center, increasing its colocalization with E2F1 in the nucleolus (Fig. 7A, right lower panel). Consistently, CoIP revealed that ADR exposure promoted the interaction between E2F1 and p14ARF in cells (Fig. 7B). To investigate the effect of p14ARF on E2F1-regulated rRNA transcription, we co-transfected E2F1 and/or p14ARF expression vector (Fig. 7C) with pIRES-RP-Luc reporter vector into H1299. Reporter enzyme assays revealed that overexpression of E2F1 promoted the activation of the rRNA promoter (Con vs. T/E2F1, P = 0.0238), whereas overexpression of p14ARF repressed E2F1-dependent activation of the promoter (T/E2F1 vs. T/E2F1/p14ARF, P = 0.0355) (Fig. 7D). Furthermore, ChIP assays showed the binding of E2F1 to the proximal site in the rRNA promoter was markedly blocked when p14ARF was overexpressed (Fig. 7E). These data indicate that increased interaction between p14ARF and E2F1 in the nucleolus may be one reason for repression of Pol I transcription activation under nucleolar stress induced by DNA damage agents.

Figure 7. p14ARF–E2F1 interaction blocks recruitment of E2F1 to rRNA promoter. (A) The co-localization of E2F1 and p14ARF in the nucleolus upon DNA damage. H1299 was exposed to ADR for 6 h, followed by immunofluorescence labeling of nuclei (blue), E2F1 (green), and p14ARF (red). Act D exposure as a positive control. The right panel shows the enlargement of nucleolar photographs. Scale bar, 10 μm. (B) Increased p14ARF-E2F1 interaction in co-immunoprecipitation assay. Cell exposure was described as in (A), and immunoprecipitation was performed using anti-p14ARF antibody, followed by immunoblot for E2F1 and p14ARF. (CandD) Repression of E2F1-dependent rRNA promoter activation by p14ARF. H1299 was co-transfected with pIRES-RP-Luc reporter vector, pcDNA3.0-E2F1, and/or pcDNA3.0-ARF plasmids. After 24 h transfection, western blotting for detecting E2F1 and p14ARF (C) and reporter enzyme assay (D) were performed. The enzyme activity in pIRES–Luc transfected cells is normalized to “1”. Data represent mean ± SD (n = 3). (E) Blockage of E2F1 binding to the rRNA promoter by p14ARF. H1299 cells were transfected by pcDNA3.0-ARF, followed by ChIP assay as described in Figure 6E. Data represent mean ± SD (n = 3).

Increased nucleolar accumulation of E2F1 couples with S phase

To illustrate the correlation of increased nucleolar E2F1 with cell phases under DNA damage stress, the cell cycle kinetics was analyzed by flow cytometry (Fig. 8A). Exposure of H1299 cells to ADR for 6 h caused an increase in S subpopulation from approximately 34% to 47% and a decrease in G₁ subpopulation from 42% to 26%, while G₂/M cells did not have any changes, compared with 6 h control (Fig. 8B). The increase of S cells by ADR could be eliminated by caffeine pretreatment, probably due to its ability to reverse S and G₂ checkpoint responses. After 12 h exposure, S subpopulation was still kept approximately 47%, but G₁ subpopulation continually decreased to 8%; G₂/M cells increased from approximately 17% to 43%, compared with 12 h control (Fig. 8C). Increased duration of ADR exposure to 24 h increased G₂/M subpopulation from 9% to more than 70%, while S and G₁ cells decreased greatly (Fig. 8D). These data indicate that ATM-mediated accumulation of E2F1 in the nucleolus couples with S phase in response to DNA damage.

Figure 8. Increased nucleolar E2F1 occurs at S phase upon DNA damage. H1299 cells were exposed to ADR or Act D for 6, 12, and 24 h under caffeine pretreatment or un-pretreatment condition. The cell cycle kinetics was analyzed by flow cytometry. (A) One representative flow cytometry analysis of the cell cycle. (B–D) Histograms showing the percentage of cell subpopulations in G₁, S, and G₂/M phases. Data represent mean ± SE (n = 3).

Discussion

In this study, we demonstrated that early accumulation of E2F1 in the nucleolus was an ATM-dependent and a common event in DNA damage. Increased nucleolar E2F1 was sequestered by p14ARF, resulting in repressing E2F1-regulated rRNA transcription. Furthermore, the accumulation of E2F1 in the nucleolus was coupled with S phase. Our data suggest that ATM-dependent E2F1 accumulation in the nucleolus is a characteristic feature of nucleolar stress in early response to DNA damage. Alternatively, early accumulation of E2F1 in the nucleolus is an indicator for nucleolar stress and a component of ATM pathway. The early accumulation of E2F1 in the nucleolus is presumably a dispersed form of ATM-induced E2F1, which elastically buffers ATM-upregulated E2F1 in the nucleoplasm, thereby coordinates the diversifying mechanisms of E2F1 acting in cell cycle progression, apoptosis, and, possibly, DNA repair in early response to DNA damage.

Most genotoxic agents perturb nucleolar structure and function.^{10,13-15,20-23,35} The concentrations of drugs are correlated with the loss of nucleolar function.³⁵ Ultrastructural studies reveal that ADR perturbs nucleolar structure in a time-dependent manner: the first clear nucleolar lesions were observed at ~2–3 h after ADR (~1.0 μM) exposure; nucleoli further increased in compactness and segregation afterwards.^14,15 We found that the accumulation of E2F1 in the nucleoli was markedly increased at 6 h after ADR (1 μM) exposure (Fig. 1). The nucleoli were full of E2F1 that was highly condensed to form conspicuous clusters (Figs. 1E and 7A). The formation of early nucleolar E2F1 should be associated with the perturbation of nucleolar structure. We also detected the repression of rRNA transcription initiation at 6 h after ADR exposure (Fig. 6B and C). In fact, the structural and functional perturbation of nucleoli is actually earlier than 6 h.^14,15 Based on our data from western blot (Figs. 1A and 3A, left panel), E2F1 protein was presumably induced within ~4–6 h. This can explain why the accumulation of E2F1 in the nucleolus was a little later than functional and structural perturbation. The increased nucleolar E2F1 within ~6–12 h corresponded to the changes in E2F1 protein levels that peaked within ~6–12 h and declined at 24 h after ADR exposure (Fig. 1A). In addition to H1299 (p53-null), increased nucleolar E2F1 was also detectable in A549 (p53-wt), 2BS (p53-wt), U2OS (p53-wt), Saos2 (p53-null), and HCT116-p53^+/+ and -p53^−/− cells exposed to ADR, etoposide (VP16), or 8-Cl⁻Ado (data not shown). We thus conclude that the accumulation of nucleolar E2F1 is a common event under genotoxic stresses, which is associated with the perturbation of nucleolar structure and function, indicating nucleolar stress.

The acute effect of DNA damage often triggers cell cycle arrest or cell death, which is principally regulated by ATM/ATR kinases.⁷ The activities of ATM/ATR and CHK2/1 promote p53 phosphorylation/activation, inducing cell cycle arrest or apoptosis. Likewise, ATM/ATR and CHK2 induce E2F1 phosphorylation/activation.^6,8 We showed that ATM, but not ATR, CHK1/2, ERK/MAPK, JNK, Wnt, and HDAC, was required for the accumulation of E2F1 in the nucleolus in early DNA damage (Figs. 2–4; Fig. S2), and S31-phosphorylated E2F1 played a critical role (Fig. 5). This is probably because the activation of ATM kinase is an early event, but others are downstream effectors of ATM or irrelevant to phosphorylation of E2F1 at S31. The ATM dependency of E2F1 accumulation in the nucleolus was further evidenced in Atm^wt and Atm^mut MEFs (Fig. 3). We therefore conclude that the accumulation of E2F1 in the nucleolus is ATM-dependent, which means that the early accumulation of E2F1 in the nucleolus is a fraction of ATM signal pathway in response to DNA damage.

DNA damage activates transcriptional and nontranscriptional functions of E2F1 through phosphorylation by ATM.^1,2,6 The high levels or modifications of E2F-1 protein may transactivate proapopotic genes, such as p73, Apaf-1, and caspases independently of p53 and cooperates with p53 through transactivation of ARF.^1,2 E2F1 may also activate expression of Chk1, ATM, DNA replication factor C (RFC), and PCNA. Furthermore, E2F1 plays additional role in DNA damage repair and replication control through interaction with NBS1, BRCA1, and Top BP1 in response to DNA damage.^1,2,4,7 In addition to DNA damage repair, DNA replication, and apoptosis, E2F1 may also mediate DNA DSBs even in the absence of exogenous DSB stimulus.³⁶ Our recent work showed that E2F1-induced DNA DSBs, in particular mitotic DNA damage, promoted drug-induced mitotic cell death.³⁷ Thus, understanding how the diverse mechanisms of E2F1 act in regulating cell death/survival and DNA damage/repair coordinate in genotoxic stresses will provide insights into tumor surveillance and limiting tumor progression. The size of nucleoli occupies 25% of the total nuclear volume in cells.⁹ We found that the nucleoli were full of condensed E2F1 clusters (Figs. 1E and 7A), and the numbers of cells with those nucleoli were increased to over 70~80% (Figs. 1D, 2C, 2G, 3C, and 4B). We thus suppose that the early nucleolar accumulation of E2F1 might be a dispersed form of activated E2F1, which elastically buffers the elevation of ATM-activated E2F1 and prevents its excess in the nucleoplasm, thereby keeping the balance between suppressive and oncogenic effects of E2F1 and leading development of injured cells in a right way. First, the early accumulation of E2F1 in the nucleolus and a moderate amount of E2F1 in the nucleoplasm may presumably decide the gene selectivity and allow the genes responsible for cell cycle arrest, DNA repair, or apoptosis turning on (and sometimes off) at particular times, depending upon their promoter affinity for E2F1 or other transcription factors (for example, p53) and upon the interaction between E2F1 and its associated proteins, such as p53, Sp1, NF-Y, and so on. Second, the nucleolar accumulation of E2F1 may also regulate timely interaction of E2F1 with DNA repair proteins such as NBS1, BRCA1, and Top BP1 in the nucleoplasm, promoting DNA damage repair. Third, nucleolar accumulation of E2F1 may benefit p53 functioning in response to DNA damage. p53 and E2F1 can control each other's transcription activities through their interaction.³⁸ E2F1 may block p53-mediated transactivation.³⁹ We recently showed that DNA damage increased the interaction between p53 and E2F1 in the nuclei, leading to repression of E2F1-dependent PLK1 gene by p53.⁴⁰ Therefore, the early nucleolar accumulation of E2F1 might be advantageous to the function of p53 in the nucleoplasm, induce cell cycle arrest or apoptosis, thereby maintaining genomic stability. Together, the early nucleolar accumulation of E2F1 may function as a buffer to produce a cushioning effect on the responsive elevation of E2F1 in the early response to DNA damage, which coordinates the diversifying mechanisms of E2F1 acts in controlling cell death/survival, cell cycle arrest, and DNA repair in the acute response to DNA damage.

ADR as a topoisomerase II inhibitor can induce DNA DSBs and perturb nucleolar structure.^14,15,28 DNA damage results in a transient repression of rRNA transcription and a temporary cessation of DNA replication.²⁵ We found that pretreatment of cells with the specific ATM inhibitor KU55933 could abrogate ADR-inhibited transcription initiation of the rRNA gene (Fig. 6B). We also found that the early nucleolar accumulation of E2F1 could be abrogated by KU55933 (Fig. 2F). A most recent study shows that KU55933 may inhibit cancer cell proliferation by inducing cell cycle arrest at G₁ phase through blocking ATM-activated AKT.⁴¹ This can partly explain why KU55933 can abrogate the inhibitory effects of ADR on rRNA transcription and early accumulation of E2F1 in the nucleolus. Therefore, our data support the notion that inhibition of rRNA transcription during DNA damage is mediated by the ATM repair pathway.²⁵

ARF controls ribosome biogenesis and cell proliferation. The ability of p14ARF to inhibit HDM2 plays a primary role in regulating p53 stability in non-transformed cells of humans.^20,42 Under genotoxic stresses, p14ARF activates ATM/ATR/CHK signaling pathways, inducing p53-independent cell cycle arrest or apoptosis.⁴³ Therefore, ARF, as an upstream regulator of the ATM/ATR signaling pathways, might act as a sensor of damaged cells. Nucleolar transcription is inhibited under DNA damage stress, in which p14ARF plays a crucial role. p14ARF controls ribosome biogenesis by retarding the processes of early 47S–45S and 32S rRNA precursors through interaction with B23.^17,18 p14ARF interacts with UBF and inhibits its recruitment to Pol I initiation complex, suppressing rRNA transcription.¹⁹ Since ATM kinase activity is required to inhibit Pol I transcription,²⁵ p14ARF may also influence Pol I transcription through Tip60-trigged ATM signaling cascade.⁴⁴ A previous report³³ and our work³⁴ show that p14ARF interacts with E2F1 to inhibit its transcription activity. E2F1 can enhance rRNA promoter activity through binding the proximal site.²⁶ We found that ADR exposure increased the levels of E2F1 and p14ARF proteins (Fig. 1A), and their interaction (Fig. 7B) led to increased colocalization of both molecules in the nucleolus (Fig. 7A). Furthermore, the binding of E2F1 to the proximal site in the rRNA promoter was markedly blocked upon ADR exposure (Fig. 6F) or p14ARF overexpression (Fig. 7E). These data indicate that p14ARF may directly block E2F1-activated Pol I transcription through association with nucleolar E2F1 upon DNA damage. It is also possible that increased E2F1 accumulation in the nucleolus might interfere with Pol I entry to the nucleolus to assemble the initiation complex. In addition, high levels of E2F1 in nucleoli may also repress rRNA transcription.²⁶ The PI3K mammalian target of rapamycin (mTOR, also known as RAFT1 or FRAP) has been shown to regulate Pol I transcription by modulating the activity of TIF-IA, and inhibition of mTOR signaling by rapamycin inactivates TIF-IA and impairs transcription–initiation complex formation.⁴⁵ Rapamycin may also prevent the formation of ATM-phosphorylated H2AX (γ-H2AX), thereby suppressing mTOR-dependent pseudo-DNA damage response in senescent cells.⁴⁶ These findings support the notion that mTOR signaling is involved in cell stresses including ribosomal stress.¹² The mTOR kinase therefore may cooperate with E2F1 under ribosomal stress. The mechanisms of rRNA transcription and its regulation are complicated. We cannot exclude other possibilities involved in inhibiting of rRNA transcription under genotoxic stress.

In conclusion, the early accumulation of E2F1 in the nucleolus indicates nucleolar stress and is a component of ATM pathway in response to DNA damage. ATM-dependent E2F1 accumulation in the nucleolus is perhaps significant in coordinating the diversifying mechanisms of E2F1 acts in cell cycle progression, apoptosis, and DNA repair in acute response to DNA damage.

Materials and Methods

Cell culture and chemicals

Human lung cancer H1299 (p53-null) cells were from ATCC; primary mouse embryo fibroblasts (MEFs) and MEFs Atm^mut were a gift from Dr Hui Zhang (Beijing Institute of Genomics, Chinese Academy of Sciences). Cells were cultured in DMEM medium supplemented with 10% fetal bovine serum (GIBICO BRL), 100 U/ml penicillin, and 100 mg/ml streptomycin, and grown at 37 °C with 5% CO₂.

Adriamycin (Sigma) (1 μM) and actinomycin D (Sigma) (10 nM) exposure was performed 24 h after placed. ATM inhibitor KU55933 was from KuDOS Pharmaceuticals Ltd; ATR/ATM inhibitor caffeine from Sigma; CHK1 inhibitor Gö6976 from Calbiochem; ERK and JNK inhibitors PD98059 and SP600125 were provided by Dr Chun-Yan Zhou (Peking University Health Science Center); lithium chloride (LiCl) from Sigma. Drug treatments were described in the “Results” section.

Constructs and transfection

E2F1 and E2F1 S31A mutant genes were released from pcDNA3.0-HA-E2F1 and pcDNA3.0-HA-E2F1-S31A (kindly provided by Dr Joseph R Nevins, Duke University Medical Center) by digestion with Hind III and EcoR I, and inserted into pEGFP-C3 vector to generate pEGFP-C3-E2F1 and pEGFP-C3-E2F1-S31A. The internal ribosome entry site (IRES)-containing pIRES-Luc was generated as described previously.⁴⁷ The human rRNA promoter sequence (−410/+314) was obtained by PCR with genomic DNA from 2BS cells as template and inserted into pIRES-Luc between BglII and KpnI sites to generate pIRES-RP-Luc. pRL-TK (Promega) as internal control of luciferase assay. DNA transfection was performed using MegaTran 1.0 (ORIGENE) following by the manufacturer.

Isolation of the nucleolus and preparation of nucleolar extracts

Preparation of nucleolar, nucleoplasmic, and cytoplasmic extracts was performed as described previously.⁴⁸ The final buffer for each extract was RIPA (50 mM Tris pH 7.5, 150 mM NaCl, 1% NP-40, 0.5% deoxycholate, 1 mM EDTA, and protease inhibitors). Protein concentrations were measured using BCA protein assay kit (Pierce).

Luciferase reporter assay

Firefly and Renilla luciferase activities were assayed using Dual-Luciferase^® Reporter Assay System (Promega). Firefly luciferase activities were verified for transfection efficiencies as computed relative to Renilla luciferase activities.

Western blotting

Ten or 30 µg of cellular fraction or total protein were subjected to SDS-PAGE, transferred onto nitrocellulose membranes (Pall Life Sciences), and probed with specific antibodies (1:1000 to 1:10 000) for E2F1 (#3742), RB (#9309), ATM (#2873), S1981-phospho-ATM (#4526)(Cell Signaling), p14ARF (MAB3782, Millipore), α-Tubulin (PM054, MBL), Lamin B1 (6581-1, Epitomics), B23 (MAB4500, Millipore), Chk2 (#3440, Cell Signaling), GFP-Tag (KM8009, Tianjin Sungene Biotech), respectively. Antibodies SAPK/JNK (#9252), phospho-SAPK/JNK (Thr183/Tyr185; #9251), p44/42 MAPK(ERK1/2; #9102), and phospho-p44/42 MAPK (Erk1/2) (Thr202/Tyr204; #9101) were from Cell Signaling. The secondary antibody was IRDye 800CW anti-rabbit IgG (926-32211) or IRDye 800 CW anti-mouse IgG (926–32210) (LI-COR). The blots were washed and imaged using an Odyssey infrared imaging system (LI-COR).

Immunofluorescence staining

As described previously,⁴⁹ cells were fixed with 4% formaldehyde in PBS at room temperature for 15 min and permeabilized with 0.5% Triton X-100 in PBS (PBST) for 30 min at room temperature. Cells were blocked in 5% BSA and 0.2% Triton X-100 in PBS for 1 h at room temperature. For labeling, fixed cell were incubated at 4 °C overnight with specific antibodies against to E2F1 (sc-193, Santa Cruz), RB (#9309), Chk2 (#3440), Fibrillarin (#2639) (Cell Signaling), B23 (MAB4500, Millipore). Then, cells were incubated with Rhodamine-conjugated goat anti-mouse IgG (1:200) or Alexa Fluor 488-conjugated goat anti-rabbit IgG (1:1000) in the blocking solution at room temperature for 1 h. After 3 washes, coverslips were mounted using 5 mg/ml Hoechst 33342 (Molecular Probes). Laser confocal microscopy was performed at room temperature using OLYMPAS.

Chromatin immunoprecipitation (ChIP) and coimmunoprecipitation (CoIP)

For ChIP assays, cells were grown in 10-cm dishes, cross-linked with 1% formaldehyde, and then lysed and sonicated to shear DNA to 200–1000 bp in length. Chromatin was diluted in dilution buffer (0.01% SDS,1% Triton X-100, 2 mM EDTA, 20 mM Tris, 150 mM NaCl) and precleared for 1 h. Then, 2 μg anti-E2F1 antibody (#3742) or IgG (Santa Cruz Biotechnology Inc) was added to precipitate the chromatin. DNA samples were purified using Wizard PCR Preps DNA Purification System (Promega). Real-time quantitative PCR amplification was performed using specific primers for the human rDNA promoter region from −126 to +20: (sense) 5′-GTTTTTGGGG ACAGGTGT-3′; (antisense) 5′-CCAGAGGACA GCGTGTCAGC A-3′.²⁶ Real-time qPCR was performed using ABI Prism 7500 Real-time PCR system (Applied Biosystems). For CoIP studies, cells were washed and scraped in PBS, then suspended in RIPA buffer for 30 min. After precleared, protein concentrations were measured using BCA protein assay kit (Pierce). Two micrograms of p14ARF (MAB3782) or IgG was added at 4 °C overnight. Protein G beads were then added, and the sample was incubated at 4 °C for 2 h and washed 3 times with ice-cold RIPA buffer. Beads were eluted with SDS sample buffer followed by western blotting.

Metabolic labeling of nascent rRNA

Metabolic labeling of nascent rRNA was performed as described⁵⁰ with modifications. H1299 cells, exposed to ADR for 6 h, were incubated in phosphate-free DMEM/10% FBS for at least 30 min and then incubated for ~0.5–4 h in the presence 10 mCi/ml ³²P-orthophosphate. The metabolic labeling medium was then removed, and cells were further cultivated for 4 h in DMEM/10% FBS. Total RNA was isolated using Trizol Regent (Invitrogen), and 1.5 μg of total RNA was separated on a 1% agarose formaldehyde gel. The gel was then dried on a gel drier connected to a vacuum pump for 3 h at 80 °C. Metabolically labeled RNA was visualized by autoradiography.

Northern blotting

As described previously,⁵⁰ total RNAs were isolated using Trizol Regent (Invitrogen). Five micrograms of total RNAs were separated on a 1% agarose–formaldehyde gel and blotted on Hybond N⁺ membranes (Amersham). The following DNA oligonucleotides were 5′ end-labeled with Digoxigenin (Dig-labeled Kit, Mylab China): ITS-1: 5′-CCTCCGCGCC GGAACGCGCT AGGTACCTGG ACGGCGGGGG GGCGGACG-3′; ITS-2: 5′-GCGGCGGCAA GAGGAGGGCG GACGCCGCCG GGTCTGCGCT TAGGGGGA-3′. Membranes were pre-incubated with Church buffer at 65 °C for 1 h, and the 5′Digoxigenin probes were then added overnight. Membranes were washed in 1 × SSC and 0.2 × SSC.

Pyronin Y-labeled RNA

In the presence of Hoechst, Pyronin Y reaction with DNA is blocked, and Pyronin Y stains RNA only. Cells were grown in a 6-well plate. Following ADR exposure and staining with Hoechst 33342, 5 μl of Pyronin Y (100 μg/ml; Sigma) was added directly to the cells and incubated at 37 °C for a further 15 min. Laser confocal microscopy was performed at room temperature using OLYMPAS.

Flow cytometry analysis

Cells were grown in a 6-well plate. After harvested cells were washed twice in ice-cold PBS and fixed in ice-cold 70% ethanol for 30 min or overnight at 4 °C. Cells were washed in PBS and digested with DNase-free RNase A (50 u/ml) at 37 °C for 30 min. Before flow cytometry analysis, cells were resuspended in 500 μl propidium iodide (10 μg/ml; Sigma) for DNA staining. A FACScan (Becton Dickinson) was used to analyze cellular DNA content. For cell cycle analysis, computer programs CELLQuest and ModFit LT 2.0ep for power were used.

Statistical analysis

Student t test and Wilcoxon rank-sum test were used for statistical analysis. Statistical significance was defined by a 2-tailed P value of 0.05.